COLLEGE OF EUROPE
NATOLIN CAMPUS
EUROPEAN INTERDISCIPLINARY STUDIES
How LLMs can help combat corruption in Ukraine?
Academic Year: 2023-2024
Statutory Declaration
I hereby declare that this thesis has been written by myself without any external unauthorized help, that it has been neither submitted to any institution for evaluation nor previously published in its entirety or in parts. Any parts, words, or ideas, of the thesis, however limited, and including tables, graphs, maps etc., which are quoted from or based on other sources, have been acknowledged as such without exception.
Moreover, I have also taken note of and accepted the College rules with regard to plagiarism (Section 4.4 of the College study regulations).
Déclaration sur l’honneur
Je déclare sur l’honneur que ce mémoire a été écrit de ma main, sans aide extérieure non autorisée, qu’il n’a été déposé auparavant dans aucune autre institution pour évaluation, et qu’il n’a jamais été publié, dans sa totalité ou en partie. Toutes parties, mots ou idées, aussi limités soient-ils, y compris des tableaux, graphiques, cartes etc. qui sont empruntés ou qui font référence à d’autres sources bibliographiques sont présentés comme tels, sans exception aucune.
Je déclare également avoir pris note et accepté les règles relatives au plagiat (section 4.4 du règlement d’études du Collège).
Keywords: Artificial Intelligence, LLMs, Corruption, Ukraine
Wordcount: 15,233
Author: Artur Willoński
Abstract
This master thesis investigates the innovative application of Artificial Intelligence, particularly Large Language Models (LLMs), as help to decrease corruption in Ukraine. Investigating beyond theoretical possibilities, the study aims to uncover practical applications that could revolutionize anti-corruption efforts in the country. Initially, the investigation focuses on exploring the technological capabilities of LLMs, including an examination of their strengths, weaknesses, opportunities, and threats. Adopting an Environmental, Social, and Governance (ESG) perspective, the study assesses the potential impact of sustainability initiatives in organizational culture and attitudes that might help introducing LLMs within administration to decrease corruption. Disparities between administrative practices in the public and private sectors are examined to identify transferable strategies conducive to effective corruption mitigation. Additionally, the study explores corruption from psychological and sociological perspectives, elucidating its underlying causes and manifestations to better address the problem and craft LLMs – based solutions. Finally, the research investigates the potential of LLMs as a preventative measure against corrupt practices, aiming to provide insights and recommendations for policymakers, practitioners, and stakeholders involved in anti-corruption efforts in Ukraine.
Contents
Introduction
Corruption remains a chronic problem in Ukraine's administration. It threatens international aid, economic development and social well-being. Despite numerous reforms over the years, corruption remains one of the most dangerous threats to the country's statehood. Ukraine sees integration with the EU as the primary social aspiration of Ukrainian society, and strategic goal. However, enlargement is accompanied by a green transformation in which the EU plays a leading role. As a result, in order to meet European standards and be accepted into the European family, Ukraine, does not only have to battle the corruption, but also introduce sustainable practices. One of the most important mechanisms for managing the sustainability efforts is environmental, social and governance (ESG) reporting. This research explores how the introduction of Artificial Intelligence (AI), in particular Large Language Models (LLMs), can influence the historical period of green transition and Ukrainian enlargement. Specifically, the focus is on how LLMs can assist in the fight against corruption through the introduction, automation and simplification of an ESG reporting process.
The aim of this research is not only to contribute to the academic discourse, but also to provide actionable insights that can inform policy formulation, drive institutional reform and catalyse positive change within the Ukrainian administration. It attempts to push the boundaries of knowledge and advance the frontiers of anti-corruption efforts in Ukraine through nuanced exploration of theoretical frameworks, empirical evidence, and real-world case studies.
Although there are also efforts to introduce AI into the public sector, the current debate on LLMs focuses primarily on the private sector. European policies such as FitFor55 and the Strategic Goals of the European Union directly integrate sustainability practices into countries’ development plans and challenge the current status quo. These efforts, manifested in various ESG frameworks, represent a significant shift in organizational culture, challenging and reshaping numerous practices within the administration itself. Environmental, Social, and Governance frameworks serve as catalysts for promoting transparency, honesty, and fair management. This substantial change in organizational culture provides a favourable environment for introducing technological innovations. Therefore, it is an optimal time to explore how LLMs can align with sustainability initiatives, catalyse change and help the fight against corruption.
To fully understand the landscape, it's crucial to investigate the interplay between ESG principles, institutional organizational culture, and their alignment with ESG requirements, including the concept of ESG reporting. The ultimate benchmark lies in observing an increase in LLM utilization coupled with a decrease in corruption levels. Based on this criterion, we can identify domains where LLM applications hold promise and proceed to showcase theoretical assessments and applicable cases aimed at combating corruption. We'll explore whether it's feasible or not, considering the complexity of the context. We'll also figure out what conditions need to be met for this technology to be viable, how it could be introduced, and where it might be beneficial or risky. Finally, we'll establish the SWOT analysis for using this technology in this specific situation of fighting corruption in Ukraine in 2024.
What are the LLMs and what is their role in Anti-Corruption efforts?
This section aims to introduce some technical terms that provide context for using LLMs and elucidate why they might be regarded as breakthrough technology. A bullet point summary sentence will be written after each small section of technological explanations. To ensure clarity and coherencies Language Models (LLMs) are advanced systems that understand and generate human-like text. In a broad sense, they are functions that use mathematical principles to predict what words will come next in a sentence. Transformers, a specialized class of Large Language Models (LLMs) introduced in 2017, leverage a mechanism known as "attention" to enhance contextual understanding. Similar to how humans focus on specific words or phrases to understand the overall meaning of a sentence, attention in Transformers helps prioritize certain parts of a text to grasp its full context. These models are versatile – they can mimic how people talk, change their writing style, and even understand 1 One of the first papers describing the use of transformers revealed its significant improvements in translation quality and computational efficiency. Evaluations on translation tasks, such as English-to-German and English-to-French, have shown superior performance compared to previous2.
Additionally, the Transformer's versatility extends beyond translation, demonstrating its effectiveness in various natural language processing tasks. Overall, the Transformer represents a paradigm shift in sequence transduction models, offering streamlined architectures that excel in accuracy, speed, and versatility, making it a pivotal advancement in the field of machine learning and language processing. LLMs are a result of Recurrent neural networks (RNNs), including long short-term memory (LSTM) and gated recurrent neural networks. Abovementioned, have long been the go-to approach for tasks like language modelling and machine translation. The way these models process sequences in a step-by-step process, generating hidden states at each position based on the previous state and input.
In other words, until now, the algorithms could only look at the previous word generated, so their sequential nature made more prolonged analysis challenging. Therefore, attention mechanisms have emerged to address this limitation, allowing models to capture dependencies across sequences without being restricted by distance. This means that no matter how far the previous text is, it can be accessed equally fast, making the process incomparably more efficient.
Traditional methods relied on recurrent networks, analysing a text word by word in the realm of language understanding and translation. While effective, this process could be time-consuming as it progressed sequentially through each word. Therefore, the Transformer model takes this concept further by discarding the recurrent networks entirely, embracing attention as its sole mechanism. This architectural shift to Transformers enables the model to comprehend the entire text simultaneously, resulting in significantly faster and more accurate translations. With just twelve hours of training on eight GPUs, the Transformer achieves remarkable performance in translation tasks, representing a substantial leap forward in language processing capabilities. It's close3. Nonetheless, the Transformer isn't limited to translation tasks alone. It can also handle complex language processing tasks like sentence parsing with remarkable proficiency, even without extensive task-specific training. This versatility positions the Transformer as more than just a translation tool—it's a comprehensive language processing solution capable of tackling diverse text based 4.
Notably, the first publicly available LLM with the aforementioned architecture Open AI’s Chat GPT 3.5 holds the record for the fastest growing consumer application to reach 100 million monthly active users, achieving this feat in just two months (estimated January 2023). For comparison, TikTok: 9 months Instagram: 2.5 years Netflix: 3.5 years (as a subscription service)5. The surge in ChatGPT's usage extends beyond mere novelty or entertainment value, transcending into professional domains. Users leverage ChatGPT's capabilities for diverse tasks, ranging from content creation to software development. Notably, the platform's versatility has found utility in programming and software development, as evidenced by data from Similar Web. Testimonials from industry practitioners underscore its transformative impact on productivity and creativity, with individuals incorporating it into various workflows to ideate, draft code, and generate content6.
How Can LLMs Revolutionize Anti-Corruption Efforts?
To understand how Large Language Models (LLMs) can revolutionize the anti-corruption efforts in Ukraine, we first need to place the technology itself in a context. LLMs hold a key position within the artificial intelligence landscape, primarily situated within the domain of Natural Language Processing (NLP). These models represent a sophisticated fusion of machine learning (ML) and deep learning (DL) techniques, particularly leveraging the abovementioned Attention mechanism in transformer-based neural network architectures.
LLMs are rooted in NLP but exhibit versatility across a wide array of AI applications, including virtual assistants, chatbots, content generation systems, sentiment analysis tools, and machine translation systems. Thaks to language understanding capabilities they can be perceived as a driving force for innovation by facilitating more natural and intuitive human-computer interactions7. Additionally, LLMs serve as accessible tools for developers, researchers, and businesses, democratizing access to advanced NLP technologies and fostering broader participation within the global internet community. Thus, LLMs occupy a central and dynamic role in advancing the boundaries of artificial intelligence, bridging the gap between human language and machine intelligence.
LLMs can be observed as revolutionary, for their capabilities to converge. That is why current research emphasizes the importance of identifying the specific type of corruption before launching anti-corruption initiatives. It is because, for instance, addressing embezzlement within the political sector differs significantly from tackling bribery schemes in law enforcement8.
AI-Recent breakthroughs in machine learning primarily drive AI-based anti-corruption efforts, nonetheless, for all AI-based technologies their most important strategy is their autonomous learning capability. Unlike traditional programming where the course of action is predefined by programmers, algorithms can autonomously solve a problem without any human intervention in a real-time. As the public administration becomes increasingly digital, with e-government initiatives, open data programs, and citizen-driven crowdsourcing efforts, these autonomous learning abilities are perceived as game-changing in a fight against corruption.
However, simply disclosing data is insufficient to effectively combat corruption. In fact, it requires individuals, whether prosecutors, journalists, or simply civil society actors, to first draw the insights from data and create a list of actions, that will be useful for policymakers. In other words, resource-intensive engagement needs to translate transparency into accountability. Without this involvement, data availability alone remains superficial9.
When it comes to Anti-corruption efforts, we can distinguish either a top-down or bottom-up approach. Top-down strategies involve implementing changes within a government structure through legislative reforms and procedural enhancements. In this case, AI plays a key role by streamlining processes and enhancing efficiency to evaluate road quality and detect potential embezzlement cases as demonstrated by projects like Kalsada in the Philippines.
Conversely, bottom-up approaches focus on grassroots efforts within society to address corruption, often involving civil society organizations and journalists. AI can support these actions by identifying and reporting of corrupt practices, exemplified by platforms like the Ukrainian portal Dozorro, which leverages machine learning – driven system to flag high-risk procurement tenders. It's important to note that AI's application in anti-corruption efforts differs from its usage in other domains, such as fraud detection or predictive policing, where governments typically employ AI to monitor citizens. In the anti-corruption sphere, it serves as a tool for both governmental and citizen-led oversight, empowering the public to scrutinize governmental activities and hold officials accountable.
However, ethical and privacy concerns arise with data leaks and crowdsourced reports, while social media data require careful examination to identify reliable indicators of corruption amid the noise. There are differences in accessibility between top-down and bottom-up approaches, with top-down initiatives often utilizing internal government data, whereas bottom-up efforts rely on publicly accessible information.
LLMs deployment on public data requires significant security measures and assurance that ethical questions are adequately addressed. On the map of technological regulatory initiatives, EU’s AI Act stands out as a regulatory framework aimed at mitigating potential risks associated with the proliferation of AI while concurrently fostering innovation and ensuring ethical deployment. The fact that European Parliament started the work on AI-ACT undermines the potential and danger that these recent developments hold. Ethical considerations, such as privacy violations and data reliability, are significant, highlighting the need for rigorous data validation processes and ethical safeguards.
As the AI-ACT relies on the validity and reliability of the input data, it is essential that the principle of "garbage in, garbage out" is heard and taken into account. While specific sources, such as crowdsourced data, may provide valid insights into corruption dynamics, ensuring the reliability of individual reports remains challenging. Conversely, reliance on contentious or unreliable data sources, such as facial recognition-based risk assessment tools, introduces inherent risks to the accuracy and legitimacy of AI-ACT outcomes. Therefore, while AI shows promise in enhancing anti-corruption efforts, careful attention to data sourcing, ethical considerations, and data quality is crucial to maximize effectiveness and address potential shortcomings.
A SWOT analysis of practical applications
To assess the practicality of utilizing Large Language Models (LLMs) for combating corruption, a SWOT analysis framework will be employed. In the assessment part, we will focus solely on the technology, the Ukrainian dimension will be discussed at the final section of this chapter. To evaluate the strengths, weaknesses, opportunities and threats it is necessary to delve into the historical background, as well as investigate usage scenarios. Subsequently, the findings will be consolidated in a bullet points summary.
A SWOT analysis framework will be employed to assess the practicality of utilizing Large Language Models (LLMs) for combating corruption, in the assessment part, we will focus solely on the technology, the Ukrainian dimension will be discussed at the final section of this chapter. To evaluate the strengths, weaknesses, opportunities and threats it is necessary to delve into the historical background and investigate usage scenarios. Subsequently, the findings will be consolidated in a bullet points summary.
Strengths
The emergence of statistical language models (SLMs) in the 1990s, was based on statistical learning methods useful for tasks such as information retrieval and natural language processing. Even with this, SLMs encountered difficulties, as models had to consider a vast array of linguistic elements or factors, making their analysis complex and computationally intensive. The subsequent introduction of Neural Language Models (NLMs) represented a significant advancement, utilizing multi-layer perceptron (MLPs) and recurrent neural networks (RNNs). This allowed to probabilistically model word sequences and was a beginning of a new architecture of transformers that is used until today and was discussed above. Therefore, the central strength of LLMs lies in their flexibility and adaptability to incorporate other technological innovations.
Another dimension, that can be perceived as a strength is that an effort is made to ensure that LLMs align well with human values and needs for their widespread use in real-world applications. In fact, to assess this alignment, various criteria are considered, such as helpfulness, honesty, and safety. Researchers use tasks like adversarial question answering and benchmarks like CrowS-Pairs and Winogender to evaluate LLMs' abilities to detect falsehoods and ensure harmlessness. LLMs learn based on the input from its users that is why, the more ease to use the model is, the more users it has and the better it gets, therefore this inherent feedback loop, also can be considered as a strength.
Moreover, the LLMs demonstrate the ability to receive feedback and perform actions based on environmental cues, such as generating action plans for agents in natural language. Evaluation in environments like VirtualHome and ALFRED involves executing tasks like household chores and compositional targets, with success measured through metrics like executability and correctness of generated action plans. Moreover, LLMs can leverage external tools, like search engines and calculators, to enhance their performance in specific tasks. LLMs gain access to fresh information through features like web browser plugins, expanding their utility.
Regarding usages, it is worth noting that large language models (LLMs) are being used in fields such as healthcare, education, law, finance, and scientific research.
In the legal domain, LLMs have become essential tools for various legal tasks, including document analysis, judgment prediction, and document generation. Their ability to interpret and reason with legal information has been highlighted by achievements such as performing well in simulated bar exams compared to human test-takers. Methods to optimize LLMs for tasks requiring complex legal comprehension and reasoning have also been developed. Furthermore, there is a growing trend towards developing LLMs specifically for healthcare applications. Notable examples include the Med-PaLM model, which has shown exceptional performance in tasks like medical question answering, diagnosis assistance, and medical document analysis; comparable to expert-level results on assessments like the United States Medical Licensing Examination (USMLE). The usages spread wide, including also the scholars who have explored how LLMs can be used in educational settings to facilitate collaborative learning, personalize instruction, and streamline assessment procedures10.
Strengths bullet points summary:
Weaknesses
Today’s discussion covers four main aspects of LLMs: pre-training, adaptation, utilization, and evaluation. Specifically, it stresses the importance of understanding the basic principles that support the abilities of LLMs trained without direct supervision. We still lack an understanding of how these models leverage pre-training data to address a range of real-world tasks, and this comes with the following weaknesses.
First of the weaknesses is a hallucination is a process where, Large Language Models (LLMs) sometimes produce inaccurate information that doesn't match existing sources or can't be verified. Even advanced models like ChatGPT struggle with this problem. One way to deal with it is through techniques like alignment tuning11.
Another dangerous weakness is reasoning, dangerous weakness is a reasoning inconsistency, so a process where LLMs might give the right answer, but they get there through the wrong path of reasoning. However, it happens also that they might reach the wrong conclusion even though their reasoning was correct. This inconsistency between the answer and the reasoning process can be reduced by adjusting LLMs based on feedback. Another way to tackle this is by considering different paths of reasoning together, like looking at a problem from multiple angles. This inconsistency can also manifest in a form of LLMs struggling with doing math, especially for numbers that they haven't seen much before during their initial training. One way to help with this is by using math-specific tools, that LLMs can get access to thanks to API; another helpful trick is to break down numbers into individual parts, like separating each digit into its own token. This can make it easier for LLMs to handle arithmetic tasks more accurately12. However, to leverage the scalability of LLMs, researchers are compiling extensive finance datasets for ongoing pre-training efforts, exemplified by initiatives like BloombergGPT, XuanYuan 2.0, and FinGPT. For example, the BloombergGPT initiative has shown effectiveness across diverse financial tasks while maintaining solid performance in general contexts.
Weaknesses bullet points summary:
Opportunities
Many opportunities lie in the model architecture, that is scalable, and ready to train, which might result in expenses decrease, and how quickly the models can make predictions. There is also a space for improvement in terms of establishing a data-centric infrastructure and introduction of training protocols of for LLMs optimization. Regarding, putting these models to use, a notable method is prompting, primarily through in-context learning (ICL), which is all about teaching the model how to tackle specific tasks13. It is worth to mention, that prompting, is simply asking model, right questions, and can be thought relatively easy.
Another opportunity lies in the field of psychology, where recent research has investigated the human-like traits of LLMs, such as self-awareness, theory of mind (ToM), and affective computing. The empirical evaluations of ToM, conducted using classic false-belief tasks, suggest that LLMs may possess ToM-like capabilities. Specific versions of the GPT-3.5 series demonstrate performance comparable to nine-year-old children in ToM tasks. Additionally, another line of inquiry has focused on integrating LLMs into software development, including functions such as code suggestion, code summarization, and automated program repair.
Opportunities bullet points summary:
Threats
Safety and alignment are crucial considerations, given LLMs' tendency to generate misleading or harmful outputs. In simpler terms, the model can generate a misleading or false information, which can propagate misinformation and counterfeit news online. Secondly, biases and stereotypes are a concern, as LLMs tend to learn and perpetuate biases present in their training data, potentially reinforcing harmful stereotypes and promoting discrimination. Safety is another area of concern, with LLMs sometimes generating offensive language, hate speech, or violent content, posing risks to individuals and communities. However, this is not as harmful, as the missed cancer due to model failure in recognizing the RTG scan properly. Similarly, in law and education, models might encounter, troubles obtaining sufficient data, to be trained before, achieving the required performance. In other words, in sensitive domains, the training might not be possible due to privacy concerns and insufficient data, to reach the minimum relevancy level. There is also a threat of hacker attack on the platform where the users access the model, in this case ChatGPT domain or Gemini or Pi.ai, might get accessed by 3rd party software and leak all the search making the platform, untrustworthy and undermining the further model development14.
Threats bullet points summary:
In the next chapter, we will step back and analyse the context in which LLMs could be introduced successfully.
ESG and anti-corruption efforts
ESG, short for environmental, social, and governance, serves as a framework utilized by diverse organizations and investors / citizens to assess an organization's operations and its impact on the environment, society, and internal management practices. It encompasses a process that must be complemented by behavioral development. This framework evaluates various factors including energy consumption, waste management, human rights, employee diversity, and many more. Essentially, ESG analysis aids in determining whether an entity conducts its activities in a sustainable and long-term perspective15.
While the research paper evaluating the implementation of ESG metrics within the European Union's bodies has not been found and requires further research if such a paper exist. EU continues to introduce new programs and strategic documents such as the Fit for 55 packages and EU's priorities from 2019 to 2024 emphasizing the centrality of sustainability16. Despite the ongoing Russian aggression against Ukraine, the shift in priorities towards security for the period 2024-2029 did not detract from the commitment to sustainability. This shows that the enduring nature of changes in values and gradual implementation in ESG reporting is here to stay.
While the concept of security may appear distinct from sustainability and ESG, it transcends traditional military hardware to encompass a fundamental shift in mindset. This transformation needs a reformation of the organizations governing the European Union and underlines the importance of self-belief and the courage to uphold stated values as based on the Estonian example17. Consequently, the sustainability of organizational culture, interpreted as a societal shift, and the efficient utilization of both material and human resources, remain central to ongoing discussions, particularly in the Ukrainian context.
How can ESG reporting influence anti-corruption efforts?
As we already know the ESG reporting is a practice within the basket of sustainable practices; it helps organizations demonstrate compliance with regulations, identify potential risks and opportunities, and enhance transparency, as well as accountability. ESG reporting comes down to frameworks, which provide guidance to all members of the organization.
These frameworks vary in scope and applicability and among benchmark frameworks, we can distinguish, the Carbon Disclosure Project (CDP) and Global Real Estate Sustainability Benchmark (GRESB), they require from organizations to respond to all questions and may include scoring mechanisms, however, there are also voluntary frameworks, such as the Global Reporting Initiative (GRI) and Task Force on Climate-related Financial Disclosures (TCFD), allowing organizations to select the topics they wish to report on without mandatory scoring.
The complexity of corruption, which spans from minor infractions to systemic malpractice makes crafting a relevant framework challenging19. The general complexity in public management and administration refers to a situation where various intricate factors contribute to the emergence of unforeseen challenges20, like the COVID-19 pandemic. The pandemic brought about an unprecedented overload, making traditional management practices ineffective and particularly vulnerable to misconduct. In this context, the concept of information stochasticity emerged, and caused an inherent randomness and unpredictability within the available data. Appearance of this phenomenon complicates decision-making, as decision-makers are confronted with incomplete and unreliable information. Researchers suggest that to address the complexities arising from the disruptive events e.g. COVID-19 it is essential to employ advanced technological tools and devise sophisticated strategies21.
Therefore, considering the complexity of corruption, the reporting mechanisms should be first standardized and then unified within already available ESG frameworks. In other words, beyond standardization, it is essential to recognize the contextual intricacies that influence the integration of anti-corruption measures within ESG frameworks. It is because, the implications of ESG metrics can vary significantly based on factors such as national regulatory frameworks, industry sectors, and organizational structures. Consequently, reporting standards should transcend superficial inquiries about the existence of anti-corruption policies to delve deeper into the underlying motivations behind an organization's historical decisions and the relevance of its policies within its unique operational context22.
Additionally, it is important to Expand ESG metrics, and try to include a better understanding of corruption risks. Among measurements that could be potentially included are:
how incentives are set up inside the organization,
how committed the leaders are to being honest,
how well employees understand anti-corruption measures,
how different teams work together on compliance,
how well anti-corruption policies are put into action,
how thorough corruption checks are, and how much stakeholders are involved. All above could help dive deeper into organizational culture and measure it thoroughly and assign them to benchmarks, what could help assess when the thresholds are crossed, and corruption risk is high.
So, there's a growing idea that we should include these details in our evaluations to understand how an organization deals with corruption. On the other hand, The Transparency International (2016) emphasizes the necessity for organizations to develop comprehensive anti-corruption programs that extend beyond the organizational boundaries to include agents and intermediaries. This underscores the importance of a holistic approach to combatting corruption. They suggest that maintaining an official website with transparent information, accessible in at least one international language, should be standard practice. The initiative argues that such transparency not only enhances accountability but also fosters trust among stakeholders.
In the realm of reporting practices, there is also a growing emphasis on consolidating information into a single platform. While current reporting standards, such as those by the Global Reporting Initiative (GRI), allow organizations to indicate where relevant information can be found, there is a proposal for all information to be readily available in one place on the internet.
In the same spirit, of precisely describing what to do and what not to do, to prevent bribery, organizations need to have strict rules against giving or receiving bribes, whether directly or indirectly. These rules should be part of a comprehensive anti-bribery program. This means they should also forbid facilitation payments, which are small bribes to speed up processes. Organizations should have written policies specifically addressing bribery, considering their unique characteristics and situations. Employees should receive training on these policies tailored to the organization's needs. Furthermore, organizations should have clear guidelines about giving and receiving gifts and entertainment. They should carefully check out third parties and business partners to make sure they're not involved in bribery23. This includes paying attention to the specific risks associated with the organization's activities and spot the weak point where the bribery might appear.
To make the reporting truly successful it is essential that a collective mobilization appears. It is of a crucial importance that for stakeholders, being the citizens to work together and ensure that their money is spent wisely. While there are collaborative initiatives for human rights and environmental sustainability, there isn't a comparable alliance focused specifically on integrating anti-corruption efforts into ESG investing. Yes, investing! Citizens are investing their money within taxes, so they should have a first-hand look at how this money are spent and make sure that reporting system works well. Working together not only reduces the cost for individuals but also creates opportunities for more significant impact. By driving collective action, people can encourage others involved in ESG processes to adopt integrity-cantered practices24.
In short, to effectively combat corruption, we need comprehensive strategies with multifaceted nature and diverse manifestations of our aim. It is the interconnectedness and individual behaviour that is crucial25. Effective approaches must understand the complex dynamics between corrupt actors, complicit individuals, and broader societal impacts. When designing anti-corruption strategies, it is important to create incentives for rule adherence, strengthen monitoring mechanisms, and impose penalties for corrupt behaviour26. Additionally, fostering a culture of integrity within governmental institutions is essential, the challenge is to sustain individual behaviours and facing group thinking introduce a method that will sustain and support individuals’ resistance to corruption, rather than inflict them on into the process to evade consequences. Moreover, a one-size-fits-all approach to anti-corruption is unlikely to succeed. The literature suggests that strategies tailored to specific contexts and challenges are more effective. Nonetheless, there is one thing, that is common for all anti-corruption struggles and this common denominator is an organizational culture.
The role of ESG driven organizational culture in anti-corruption efforts?
ESG-driven organizational culture is not special, it is organizational culture and doesn’t play any special role27. Nonetheless, on the sustainability wave it means that organization is focused on measuring all actions, and holding leaders and decision-makers are directly accountable for these actions before stakeholders28. Their primary objective is to ensure that processes are maintained within the framework of sustainability. While organizational culture is a well-researched topic in the context of profit generation, conflict resolution, hiring practices, and, importantly, corruption mitigation; in companies, the driving forces behind the culture of public organizations often differ and cannot be directly replicated. However, from another perspective, the underlying biochemistry behind decision-making in both private and public organizations remain the same, allowing us to draw insights from extensive business literature while bearing in mind this distinction between the private and public sectors exist. Considering all these factors, the concept of ESG-driven organizational culture in the public sector is relatively new and warrants further research.
Organizational culture is an idea of how things in a company are supposed to be done, and every organization has an idea of that. The goal of organizational culture is for all the elements that make up organizational culture to be consistent with each other and work like a well-oiled mechanism29. Although a direct causal link between organizational culture and corruption doesn’t exist, the improvements in building block of culture can mitigate corruption significantly30.
Balance between efficiency and resilience is key to maintain an organizational culture that supports the implementation of ESG. As we read in the MDTI report, there are times when, within a single organization, departments report independently of each other, piling up and complicating the entire process. It happens that it is unclear to reporters what they should or should not report. Other times there is a lack of historical data, to the missing databases need to be purchased31.
When sustainability is an additional task to manage and not a central point to all actions, new problems arise. Manipulation of indicators, economic pressure, or fear arising from the fear of revealing too much information. We see that to deal with these adversities, the solution is strong leadership and commitment at the highest level. Methodological challenges include the difficulty of defining appropriate quantitative measures for social issues, setting ambitious and realistic ESG targets.
The challenge extends beyond mere enhancement and ongoing refinement of reporting processes. It encompasses a broader issue: stakeholders themselves often lack an understanding of what ESG entails and its implications. The scope of ESG is vast, spanning numerous topics, making it easy to become overwhelmed, especially when juxtaposed with CSR (Corporate Social Responsibility), which, while similar in role, is not identical. Moreover, for those tasked with implementing these standards, the expansive legal landscape is a barrier to pave a way for next actions. Without a consensus on the definition of sustainability, moving forward becomes hard. Additionally, there's the challenge of acquiring a diverse range of expertise necessary for effective reporting—from legal professionals to technologists to environmental specialists—posing organizational and cost-related hurdles.
The reason why organizational culture might be perceived as important is that 2/10 employees are guided by their company's culture in their daily decisions. Even though, most high-level managers feel attached to their organization's culture, majority of employees do not share this feeling. Considering that this has a direct impact on the quality of candidates, the amount of talent on the team and productivity, these results are alarming.
So, what's stopping organizations from setting, aligning, unleashing and maintaining a high-performance organizational culture? Exactly that - the complexity of the entire process32.
A strong organizational culture has an impact on key performance indicators. Possible is even 50-point increase in employee engagement, a 25% increase in the number of employees, all possible within 3 years. On the other hand, within 5 years observable is an 85% increase in net income. Building an organizational culture is based on 5 key actions, and the first is to connect employees with purpose, then create a positive employee experience, and strive to maintain maximum authenticity and transparency in operations, a culture of 1-on-1 conversations and employee appreciation should also be maintained. The required alignment can be achieved when employees' goals and the company's goals move in the same direction33. Similarly, employee appreciation, can take the form of public praise, but can also be a series of informal small gestures of kindness.
The general functioning of an organization is also a continuous process of creation consisting of 2 main elements. The first is to focus on recognizing who deserves praise, so as not to overlook people who often stand in the shadows, although they do critical work; there is always a good time to recognize someone’s achievements34 and management doesn't have to wait until goals are achieved or until the quarterly summaries. Social recognition is key in this regard, with 92% of employees agreeing that when they are properly recognized for their action on behalf of the organization, they will repeat that action in the future. The second of the foundations is talking about strengths and feedback, as the report indicates - company profitability increased by 8.9% in companies where managers received feedback on their strengths35.
In a company, every employee, should be guided by the values of the company. Values, however, are tested and built into the fabric of the company over time, projects or simply conversations in the hallways between employees. Apple Inc.'s famous headquarters, designed to increase the likelihood of interaction between employees from different departments, makes profound sense as a stimulator of creativity, but its main purpose is to build a space where employees can talk and embed the company's values in their daily lives36Continuous learning is another important element, as organizational cultures are formed by employees who are constantly learning. We're talking about continuous learning on the initiative of employees, but also individualized workshops, trainings and professional development courses. Moreover, companies that invested in innovative learning solutions and were willing to increase their training budgets showed increased revenues in the previous year37.
In the public sector once introduced, ESG operates like a snowball effect, initiating a cascade of improvements across various processes. For instance, when managers undergo ESG training, they begin applying these principles in their daily operations, thereby enhancing team security and fostering an environment conducive to innovation and creative thinking. This creativity, in turn, gives rise to additional initiatives that yield better outcomes, lowering the threshold for implementing transparency measures. Once transparency measures are established and ESG-related processes permeate the organization, foreign aid38 can play a crucial role. Given that good governance is essential for attracting foreign aid, particularly in the case of Ukraine, whose reliance on foreign aid is significant, this becomes even more important.
Innovation is yet another way to sustain development with the same resources, and when it comes to public administration, the emphasis on the learning curve and innovation is assessed in a more rigid manner than in companies39. The major problem of public institutions is that they operate in a world that has changed, and they don’t adapt to evolving circumstances.40
The study analysing the Social Progress Index, finds that larger governments tend to slow down economic growth. However, when institutions are strong, the negative effect of government size is less pronounced. Also, spending on non-development areas hurts growth more than spending on development when institutions are strong. Overall, the study suggests that governments should focus on improving governance quality and keeping their size smaller to promote economic growth at the state level41. At this juncture, one might ponder how administrations can maintain their compact size. Automation serves as a prominent example of how this can be achieved; interestingly, automations are also a prominent use case for many large language models (LLMs)42. Indeed, many of the issues that arise within the realm of organizational culture might result in corruption; and at the same time, they might be mitigated through the utilization of LLMs.
Corruption
Corruption typically is described as denoting improper and typically unlawful behaviour aimed at obtaining personal or mutual gain, manifests through various forms such as bribery, extortion, and the exploitation of privileged information43. It is proven that corruption has a detrimental effect on the economy and is perceived as a sign of poor governance rather, than its cause.
Corruption often arises due to inadequate oversight by the central government, leading government agencies to engage in illicit activities. These agencies may demand bribes from private individuals or businesses seeking permits or approvals. When there is minimal regulation regarding the appointment of personnel within these agencies, they can exploit their positions to solicit bribes repeatedly. For instance, in post-Communist Russia, foreign investors frequently encountered demands for bribes from multiple agencies, discouraging them from investing in the country and adversely affecting its overall business environment.
The second primary reason for the high cost of corruption is the distortions caused by the necessary secrecy surrounding corrupt practices. Secrecy can divert investments from high-value projects like health and education towards less beneficial areas like defence and infrastructure, simply because they offer better opportunities for secret corruption. Additionally, secrecy can lead to monopolies, discourage innovation, and inhibit useful investment and growth44.
Corruption arises in environments characterized by a lack of transparency and weak oversight. It is a complex phenomenon that manifests uniquely in various contexts, yet its underlying requirement is a deficiency in honesty, transparency, and a significant lack of adherence to ethical values.
Corruption in Ukraine
To understand the corruption in Ukraine, one must examine both microeconomic factors, which are process-oriented, and macroeconomic issues prevalent in the broader society. While it's crucial to prioritize the processes, the role of individuals should not be overlooked, as they are the ones demanding reforms since the Orange Revolution and later on during the Euromaidan Protests in 2014. Despite efforts to increase transparency and curb administrative corruption, prosecuting corrupt officials remains a significant challenge. Even though new laws and anti-corruption bodies have been established, corruption continues to be pervasive in Ukraine, contributing to its reputation as one of the most openly corrupt countries in the world. The phrase "openly corrupt" suggests that corruption is widely recognized and visible across various sectors of Ukrainian society, which is a positive first step towards addressing this problem.
Nonetheless, for the change to go to the next level, the country’s elites must find a consensus to change the underlying governance rules. Historical experiences from Western Europe demonstrate that practical anti-corruption efforts require transitioning from a "limited access order," where a small group of influential individuals dominates through patronage, to an "open access order" that accommodates a broader range of interests based on unbiased institutions and social relationships. Assuming such a transformation takes approximately 50 years, with Ukraine's starting point in the 1990s, the country would currently be at the halfway mark in this process45.
Nonetheless today, Ukraine is a state where corruption is deeply ingrained and is the foundation for power dynamics among elite groups. Despite progress in reducing corruption, overall levels remain high, requiring further action. Punitive measures against corrupt officials are acknowledged but are deemed insufficient due to the established nature of the corrupt system. Simply prosecuting corrupt individuals without addressing the systemic issues underlying corruption is seen as ineffective46.
In Ukraine, corruption is present at multiple levels of government, with mid- and low-level officials often expecting incentives in the form of privileges or illegal side payments derived from state resources. This system establishes a dynamic where corruption serves as both a reward and a threat. In simple terms, these illegal small payments are acting as a binding force that holds the current system together. It is crucial to acknowledge that numerous state institutions heavily rely on these informal mechanisms, particularly in law enforcement. Abruptly dismantling these systems could lead to a rapid decline or even paralysis of these institutions. Nevertheless, there are successful examples, such as the Ministry of Health's medicine procurement system, which underwent significant reform without causing the institution to collapse entirely.
Particularly in limited access orders like Ukraine, specific forms of corruption, may paradoxically facilitate economic and societal development by bypassing bureaucratic inefficiencies. However, it warns that such corruption can have harmful consequences, including embedding informal practices that are challenging to remove and compromise the safety regulations, leading to tragic accidents. Moreover, tolerating certain forms of corruption can normalize unethical behavior and blur societal boundaries between acceptable and unacceptable conduct47.
Although, there are many initiatives to reduce opportunities for corrupt practices, challenges persist in holding corrupt officials accountable. It is essential to understand that corruption is symptomatic of deeper systemic issues in governance rather than being the sole cause. Achieving meaningful change requires opening Ukraine's political landscape to increased participation, fostering competition, and establishing robust institutions capable of upholding the rule of law.
Various sectors have witnessed significant strides as part of the anti-corruption efforts, including restructuring entities like Naftogaz and implementing reforms in administrative services, banking, law enforcement, procurement, taxation, and decentralization initiatives. These efforts have limited avenues for corruption and empowered citizens to actively hold local authorities accountable for managing public resources.
Progress in addressing corruption in Ukraine since 2014 has been mixed, with notable advancements in specific sectors contrasted by persistent challenges in others. While administrative services, banking, and law enforcement have seen improvements, critical areas such as customs, deregulation, privatization, de-monopolization, and public administration reform lag. The lack of transparency in defense spending is a particular concern, indicative of broader issues surrounding accountability within the sector.
While establishing the National Anti-Corruption Bureau represents a significant step forward, its effectiveness in prosecuting high-level corruption cases has been limited by the influence of vested interests on the judiciary. The forthcoming High Anti-Corruption Court holds promise for enhancing accountability but may encounter resistance and concerns about selective justice. Recognizing that punitive measures alone are insufficient to combat corruption comprehensively is essential. A holistic strategy must encompass systemic reforms to curb opportunities for corrupt practices and foster a competitive political and economic landscape. This includes dismantling political monopolies and incentivizing influential actors to embrace transformative norms of behaviour.
There's a notable contradiction in Ukraine: while high-level corruption is widely condemned, many people consider petty corruption acceptable. This dual mindset reflects a complex relationship between society and corruption. However, it's crucial for attitudes to change. Even small acts of corruption contribute to societal decay. To combat this, citizens must reject all forms of corruption, big or small. By promoting transparency, accountability, and ethical conduct, we can work towards a fairer and more just society48.
Understanding any society requires an examination of its informal institutions, which play a significant role in shaping both its economic and political landscapes. In the case of Ukraine, endemic corruption stands out as the most influential informal institution, posing a substantial threat to the nation, alongside Russia's efforts to undermine Ukraine's territorial integrity and sovereignty. During Viktor Yanukovych's presidency, corruption in Ukraine manifested in three primary forms: firstly, through the manipulation of natural gas trade; secondly, through Yanukovych's arbitrary allocation of major infrastructure projects; and thirdly, by using their access to public resources for personal gains. With Ukraine's recent shift towards alignment with Europe, there is optimism that the nation now has an opportunity to address and mitigate the deep-seated corruption that has plagued its economy and political system for years49.
In Ukraine, corruption takes on various forms, ranging from petty corruption to grand corruption. Petty corruption is widespread and often considered inevitable by many citizens. This acceptance is often rationalized by pointing to the involvement of high-ranking officials and oligarchs in more significant corrupt practices. Experts estimate that corruption results in substantial financial losses for the country, amounting to billions of dollars annually. On the other hand, grand corruption involves the abuse of power by individuals in influential positions, leading to significant harm to society while benefiting a select few. It thrives on informal networks connecting government officials, legislators, judicial figures, law enforcement agencies, and politically connected entities or individuals. This complex network of corruption undermines public trust and inflicts severe and widespread damage on individuals and the nation50.
"State capture" refers to the phenomenon of powerful political and economic elites gaining control over public institutions and the economy, forming a hierarchical structure that serves their interests. This issue has been identified as a significant aspect of corruption in Ukraine. Both the International Monetary Fund (IMF) and the Ukrainian government have acknowledged the strong resistance from entrenched vested interests against implementing structural reforms. Grand corruption, stemming from weaknesses in the rule of law and the pervasive influence of oligarchs, contradicts the values upheld by the European Union (EU) and poses a substantial obstacle to Ukraine's progress. Besides stifling competition and impeding economic growth, it undermines democratic processes and fosters widespread petty corruption. Moreover, investigative journalism has highlighted instances of oligarchs engaging in illicit financial activities, such as money laundering, even within the EU.
Estimates suggest that tax evasion through offshore channels significantly affects Ukraine's economy, amounting to at least one billion euros annually. Despite concerted efforts to attract foreign investment, persistent challenges such as pervasive corruption, a lack of trust in the judiciary, and market barriers persist unabated. Various empirical studies, including those conducted by the IMF, underscore a noticeable correlation between reducing corruption, economic growth, and improving citizens' well-being. Recognizing the need for substantive change, stakeholders agree on the importance of addressing the entrenched influence of vested interests as a prerequisite for fostering 51.
Corruption in Ukraine poses significant challenges to the nation's security, stability, and democratic progress. While specific sectors like healthcare have seen improvements, critical areas such as defence and energy remain plagued by secretive practices and low transparency. The acceptance of small-scale corruption may be partially attributed to bureaucratic inefficiencies, but it also reflects how elites exploit the system to perpetuate corruption and maintain a lack of clarity. This distinction between acceptable small-scale and unacceptable large-scale corruption creates a harmful dynamic where external influences are harder to spot. Moreover, the concentration of media ownership among oligarchs grants them significant influence over public discourse and societal norms.
Anti-corruption measures in Ukraine
Similarly, to corruption roots analysis, anti-corruption measures can be divided into process-oriented and macro-economic and more social measures. This means encouraging people to come together around common interests and goals, which can help create a more sustainable way of fighting corruption over time.
There are numerous measures to combat corruption in Ukraine as the first one of a more social nature is an alliance comprising reformist politicians, business leaders, and civil society actors holding the potential to drive meaningful change. Nonetheless, it requires robust support from Ukraine's international partners. Civil society organizations must enhance their capacities and extend their reach beyond significant cities to advance this attempt. At the same time, the public needs improved access to unbiased reporting through independent media outlets52. Ukraine, of 2024 is very different from Ukraine in 2014, and a lot has already been done to resist disinformation. Ukraine today is perceived as the most resilient to Russian disinformation53, nonetheless the danger remains high.
In 2014, Ukraine entered an Association Agreement (AA) with the European Union (EU), signaling a strategic alignment toward political and economic integration with European norms and standards. Unlike previous agreements, this pact did not guarantee Ukraine's eventual accession to the EU but required comprehensive reforms to enhance governance and institutional capacity. Despite the ambitious scope of reform commitments outlined in the AA, Ukraine has encountered challenges in implementing these reforms effectively. Before 2014, EU assistance to Ukraine primarily focused on raising awareness of European regulations and standards. While this aid contributed to knowledge dissemination, its impact on enhancing the functionality of Ukrainian state institutions remained limited. Following the signing of the AA, the EU intensified its support to Ukraine, aiming to facilitate the reconstruction of state institutions. This expanded assistance sought to bolster Ukraine's governance structures and administrative capabilities, aiming to align them more closely with EU norms and practices.
Several key innovations have been implemented to support Ukraine's reform efforts. These include establishing the Support Group to Ukraine (SGUA), longer and more substantial assistance programs under devolved agreements, dedicated staff positions for overseeing reforms, and extensive macro-financial assistance. These initiatives aim to enhance the effectiveness and impact of support provided to Ukraine in its reform actions. At the macro level, the Support Group to Ukraine (SGUA) has introduced a more comprehensive approach to coordinating and planning assistance, emphasizing sector-wide initiatives rather than individual projects. Additionally, the SGUA has facilitated better coordination with other international donors, enhancing the overall effectiveness of support efforts.
Furthermore, high-level EU officials in Ukraine have provided valuable expertise to support political and technical interactions with the Ukrainian government. Their involvement has contributed to the advancement of various initiatives and reforms. On a micro level, much of the EU's support comes from technical assistance projects, which focus on transferring specific knowledge and skills. While effective, these projects often have limited scope and short durations, posing challenges for institution-building efforts. This limitation is not unique to EU assistance but is nevertheless a notable factor affecting the impact of support programs54.
The European Union (EU) has actively supported anti-corruption efforts in Ukraine. However, despite these actions, grand corruption remains a formidable challenge within the country. Through bodies like the European External Action Service (EEAS) and the Commission, the EU has recognized the intricate connections between Systema stakeholders and has addressed corruption as a cross-cutting priority.
However, a comprehensive strategy for grand corruption within EU initiatives remains lacking. The EU's Anti-corruption Initiative project has primarily focused on combating corruption but has not explicitly targeted grand corruption. While some EU-funded projects include activities aimed at addressing aspects of grand corruption, such efforts have not been comprehensive. Additionally, there is a noticeable gap in addressing money laundering risks within these projects. Conditions for reform have been incorporated into EU assistance programs, yet the EU's advisory mission in Ukraine does not explicitly tackle grand corruption within its mandate. This highlights a need for more targeted and coordinated efforts to effectively combat grand bribery in Ukraine, emphasizing the importance of comprehensive strategies and robust institutional reforms55.
Furthermore, there is an urgent need to strengthen monitoring mechanisms and support the effective implementation of anti-money laundering legislation within Ukraine. This is crucial for enhancing resilience against illicit financial activities, commonly associated with grand corruption schemes. Additionally, the Commission's support for civil society initiatives and independent media is commended for their role in preventing and detecting corruption. Platforms such as ProZorro and those dedicated to monitoring politically exposed persons are particularly praised for contributing to reforms and raising public awareness about corrupt practices and abuses of power56.
Another anti-corruption step is prioritizing reforms within key institutions like the judiciary, prosecution services, and law enforcement agencies to ensure their independence and integrity. This involves implementing strict requirements to promote transparency and accountability within these sectors. Additionally, the report emphasizes the importance of assessing and improving data accuracy within the Ministry of Justice's registries, particularly concerning e-asset declarations. Addressing issues related to the unlawful takeover of businesses or property is identified as crucial for mitigating risks for investors, and steps are being taken to tackle this challenge. In terms of supporting civil society and investigative journalism, there is a call to adjust the scale of support provided, emphasizing combating corruption and reducing the influence of media owned by oligarchs.
Another interesting aspect of the public administration is the non-managerial employee selection technique. It remains outdated despite some improvements in recruitment procedures since 2018. Issues persist in the civilian workforce's salary system, including pay disparities and unclear bonus allocation procedures. The Civil Service Reform Initiative (RSP) aims to attract professionals to administrative roles by offering significantly higher salaries. Nonetheless, as salaries rise, the well-educated workforce, fluent in EU and international arena procedures, is flying out of the country, leaving it without the necessary expertise. This process undermines the efforts to build a sustainable corruption-free organizational culture and modernize public service practices. It remains to be solved how to ensure fair compensation and enhance transparency, aligning with Ukraine's broader goals for administrative reform and anti-corruption measures57.
The Green Agenda project, launched by the Stockholm Environment Institute (SEI) in collaboration with the Ukrainian government and Sida in Kyiv, aims to align national policies with the principles of the European Green Deal. It seeks to promote sustainable development and facilitate the path to EU membership. The opening event attracted many stakeholders, including government officials, international community representatives, NGOs, private sector actors, and academia. Despite Russian aggression, Ukraine remains committed to the Paris Agreement and strives for climate neutrality. Integration with the EU promises benefits such as access to a larger market, new trade opportunities, and economic growth. The Swedish Ambassador to Ukraine highlighted the potential of combining sustainable development with EU integration to create jobs and develop new industries. SEI pledged to support Ukraine in achieving its climate goals by providing tools and knowledge for informed decision-making during its reconstruction59.
In Ukraine, there is a visible surge in enterprises who are prioritizing ESG (Environmental, Social, Governance) criteria, and are ready to be listed through the ESG Transparency Index Ukraine 2020. Aligning with UN Sustainable Development Goals, companies are integrating ESG strategies into their business models. State-owned enterprises typically outperform privately-owned ones on ESG criteria. Nonetheless, The National Bank of Ukraine and the National Securities and Stock Market Commission are spearheading efforts to embed ESG factors into corporate governance systems. Transparency and disclosure of ESG information are increasingly crucial for investors and business growth in Ukraine, reflecting a shift towards sustainable practices and responsible investing60.
Numerous workshops aimed to develop a National Framework Guideline for green reconstruction. They facilitate sharing experiences and discussing the role of sustainable development standardise in green recovery. These initiatives have potential to ingrain positive practices thanks to national experts and public and private sector stakeholders and discuss integrating sustainable development into Ukrainian public policies for post-conflict rebuilding. Panellists include representatives from UNIDO, British Standards Institution, CEN and CENELEC, UAS, and Ukrainian industry associations. EU support is evident through active involvement in standardization and green rebuilding negotiations. CENELEC aids European standardization and collaborates with Ukraine on trade, sustainable development, and innovation. UAS's affiliation with CEN & CENELEC grants access to green rebuilding standards. 61.
In Ukraine, significant changes include involving civil society in senior MOD roles and establishing anti-corruption training and a comprehensive plan. This strategic document should emphasize the importance of collective leadership commitment to change, specialized anti-corruption expertise, and tailored approaches for specific sectors like defence. Overall, tackling corruption requires legal changes and a cultural shift within institutions and leadership62.
Facing challenges and experiencing difficulties, such as those encountered in efforts to combat corruption, can strengthen individuals and organizations. It's like saying that when things get tough, instead of giving up, facing those challenges, head-on can build resilience and character. In the context of anti-corruption efforts, it means that even though the journey might be challenging, enduring the difficulties and maintaining perseverance is crucial. It emphasizes the importance of having a solid and unique organizational culture that encourages resilience and determination to overcome obstacles in the fight against corruption63.
The anti-corruption struggle in Ukraine is marked by a resilience that tends to grow over time. This situation suggests that as individuals encounter and overcome obstacles, their determination to combat corruption is increasing.
Ukraine is already mitigating the media's influence on societal perceptions and implementing effective monitoring mechanisms. It is Noteworthy that Ukraine's aspiration to join the European Union is a driving force behind its commitment to anti-corruption efforts, emphasizing its dedication to institutional reform and transparency.
Sources and Methodology
This research aims to investigate the national security landscape of Ukraine, focusing on the development of LegalTech and Machine Learning Models (LLMs) and the incorporation of Environmental, Social, and Governance (ESG) requirements. Employing a literature review as the primary research method, the study seeks to contribute to the advancement of knowledge and the development of organizational culture theory64. It will draw insights from empirical studies in areas such as artificial intelligence, sustainable growth, and national security to establish correlations between these topics. By doing so, the research tries to enhance understanding of national security dynamics and promote the cultivation of an organizational culture conducive to the reconstruction efforts in Ukraine.
The research methodology uses scholarly databases like Google Scholar, JSTOR, and ScienceDirect to procure relevant academic literature, reports, and publications. Additionally, Google's Large Language Models (LLM) tool, particularly Gemini, is employed to generate prompts, which are then utilized in Google Scholar searches. Below is the table with all prompts generated and used in the research.
The synthesis of existing research is performed on organizational culture, LLMs, ESG, and their potential impact on public administration. In the next step of analysis, it is necessary to identify areas where LLMs can be most beneficial within public administration. Additionally, analysis of the challenges, opportunities, strengths, and weaknesses (SWOT analysis) of LLM is performed. Next, the assessment of ESG potential in introducing LLMs within public sector is assessed. Furthermore, research explores the potential relationship between ESG – driven organizational culture, where LLMs could be introduced with the purpose of helping anti-corruption efforts in Ukraine.
We establish a null hypothesis stating: "There is relationship between the ESG – driven organizational culture in public institutions, where LLMs were introduced and higher chances to help combat anti-corruption efforts in Ukraine."
This research relies solely on existing literature, and no primary data collection is conducted (surveys, interviews). The findings are limited by the availability and scope of existing research on the specific intersection of LLMs, ESG, organizational culture and corruption studies
Analysis
The research hypothesis does not reject the null hypothesis, suggesting a correlation between the ESG-driven organizational culture within public institutions, where LLMs were implemented, and the capacity to aid anti-corruption endeavors in Ukraine. It is crucial to note that the inclusion of the ESG parameter in the study was solely aimed at amplifying the potential impact of LLMs on anti-corruption efforts and serving as a variable that enhances the likelihood of success, given its macroeconomic significance.
Large Language Models (LLMs) offer multifaceted potential in enhancing organizational dynamics. Firstly, they can analyse extensive communication data within companies, unveiling patterns in employee sentiment, communication styles, and cultural norms. This insight can shape targeted training programs, bolstering aspects of organizational culture.
Secondly, LLMs can detect biases in internal communications or policies, fostering inclusivity. Finally, LLMs can serve as collaborative tools, generating interactive simulations for employees to refine skills like communication and conflict resolution. These applications not only cultivate empathy and understanding but also promote effective communication within the workplace, facilitating a culture of growth and harmony.
It explores how LLMs are revolutionizing AI and solving real-world problems. Despite their potential, deploying LLMs poses challenges like ethical concerns, biases, and computational demands. It is important to note that we need Techniques to enhance LLM robustness and address biases for better performance65.
Education of civil servants and civil society
One of the most well-known LLMs is ChatGPT, offers a wide range of applications in higher education, facilitating student interaction with the system and easing administrative tasks. It can support active learning through interactions, immediate feedback, and reducing administrative burdens. Integrating ChatGPT may foster a conducive learning environment, enhancing student engagement and saving costs for educational institutions. It can also provide personalized educational experiences, efficient resource utilization, and an adaptive approach consistent with constructivist theory. However, challenges arise regarding academic integrity, access inequalities, and the reliability of information delivery. Additionally, its impact on critical thinking in assessments and ensuring fairness in evaluations pose concerns. The fact that it is a privately held and widely available technology, makes it prone to hacker attacks, therefore high-risk technology for the personnel of public sector66.
As previously assessed, corruption in Ukraine may stem from inadequate education within civil society. Despite improvements following the 2022 war, which reduced susceptibility to Russian propaganda, the core issue lies within Ukraine itself. Despite the presence of platforms like Telegram and numerous independent channels, traditional media in Ukraine remains under the control of oligarchs. To support Ukrainian society, it is imperative to empower civil society with additional sources of knowledge. LLMs have demonstrated significant efficacy in educational contexts, offering personalized teaching approaches that cater to the individual learning pace and preferred communication style of each citizen.
Mindset Shift for small corruption combat
LLMs have the potential to contribute positively to the mindset shift integral to anti-corruption efforts. One of their strengths lies in their ability to be tailored to specific industries and trained on local data, aligning with the Ukrainian administration's efforts to enrich existing datasets for innovation within the country. It is noteworthy that LLMs are predominantly semi-supervised models, meaning human oversight is crucial to prevent biases during their development. This responsibility is often delegated to countries with lower labour costs, where workers, including those from the Ukrainian administration who may face insufficient remuneration, can contribute to shaping the values embedded within the model. This presents an additional financial incentive for workers and offers them the opportunity to influence the cultural norms within institutions and share them with other stakeholders in the state. Consequently, this has the potential to instil the desired culture within institutions and stimulate the necessary mindset shift for long-term success in combating small-scale corruption. However, it is important to recognize that this process is susceptible to disinformation and sensitive in its architecture. In other words, if the individuals training the models are not trustworthy, the models themselves may not be trustworthy either.
However, it is easier to identify a smaller group of people who will train the models and receive additional compensation for their efforts, and this group may be less susceptible to bribery compared to assuming that all lower-level administrative workers hold similar values without sufficient reason. It is plausible that corruption stems from low-level development, where individuals seek to increase their earnings. One potential solution is to automate the system to require a limited number of personnel, thereby allocating sufficient funds for salaries and rendering corruption economically unfeasible. Interestingly, in 2020, Russian aggression was ranked eighth out of ten top risks, deterring foreign investments. Similarly, the significant outflow of human resources from Ukraine was ranked tenth, considered the least probable occurrence67. Remarkably, four years later, these unlikely events became reality.
Data privacy is a big concern when it comes to training AI models, especially in fields like law, medicine, and education. There's also the risk of hacker attacks targeting the online platforms where these models are hosted, which could mess up their functioning. Another issue is that biased datasets can lead to AI models picking up and reinforcing stereotypes and biases, which isn't good. Plus, there's the problem of AI models sometimes generating false information, which can spread misinformation. People also have some outdated ideas about AI, which might stop them from using it effectively, especially in academic and social settings. And even experts don't fully understand how some AI training processes work, which makes it hard to optimize these models. Sometimes, AI models can give the wrong answers even if their reasoning seems right, which is frustrating. They can also struggle with math problems involving numbers they haven't seen much before. Plus, it's tough to keep AI models performing well when you add new, specialized data to them. And updating their knowledge takes a while, which slows things down.
Limitations
The research results highlighted certain limitations in the study, particularly in the thorough assessment of the role of ESG and its impact within the context of introducing LLMs. Therefore, further research is needed to explore the relevance of sustainability and ESG-driven organizational culture in combating corruption in Ukraine. Additionally, as the development of LLMs progresses each month, this research should be considered as a suggestion and guide for conducting more detailed analysis on the opportunities and challenges ahead for those considering the implementation of these systems in their respective administrative offices.
Conclusions
It is easier to identify bias and minimize the scope for corruption when LLMs are involved in monitoring processes, as their architecture allows quick retrieval of any data required. For the human analyst, it makes navigating through the big data sets easier and faster thanks to the simple ability to write text message prompts - no programming required.
With the introduction of ESG reporting, the fact that there is a large language model in place to interpret the raw data simply reduces the scope for potential corruption. Because the entire process is monitored and reported in real-time, there is less opportunity to corrupt the results.
To ensure a high level of accuracy, LLMs can be trained on specific datasets. It can then be used internally by Ukrainian administration for two-way communication with external partners, such as the European Union in the accession process. A bilateral online platform like this, could avoid translation confusion of the documents and make the process faster. It could also serve as a 24/7 online monitoring center available for all stakeholders assuring them of the necessary transparency. In a similar way to ProZorro, for internal fair tendering, but this time for the international aid process.
Open public LLMs, which are national chats, similar to the already existing telegram channels, might serve as a way to improve automations and give people job. Namely, opening the training of the model to the citizens, i.e. offering them the possibility to supervise the learning curve of the model by answering the questions, would have an additional impact on building self-awareness and fighting the small corruption culture.
The role of this individual involves the correction of grammatical errors, such as the replacement of "serbant" with "servant." This is an opportunity to create a new group and give people a job to increase the efficiency of the administration, not by giving bribes, but by helping build public online platform that will help everyone.
This framework illustrates the potential for a mindset shift in addressing small-scale corruption and highlights the utility of Large Language Models (LLMs) for democratization purposes, reinforcing civil society and combating major corruption. However, while the proposed solutions offer theoretical insights, similar implementations have already begun to emerge in the private sector. Therefore, further research is necessary to develop more detailed assessments and actionable plans.
The research concluded that the deployment of LLM presents significant challenges, including ethical concerns, biases, and computational demands. These obstacles are preventing decision makers from engaging more fully with this technology. It was found that there is a pressing need for techniques to enhance LLM robustness and address biases to improve performance. Additionally, the susceptibility of LLM training processes to disinformation highlights the importance of ensuring the trustworthiness of individuals involved in their development.
Bibliography
Rohit Madan, Mona Ashok, AI adoption and diffusion in public administration: A systematic literature review and future research agenda, Government Information Quarterly, Volume 40, Issue 1, 2023, 101774, ISSN 0740- 624X, https://doi.org/10.1016/j.giq.2022.101774. (https://www.sciencedirect.com/science/article/pii/S0740624X22001101)
Google. (n.d.). Large Language Models: OpenAI's GPT and Beyond. Retrieved from https://developers.google.com/machine-learning/resources/intro-llms?hl=fr
Vaswani, Ashish, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. "Attention Is All You Need." 2023. arXiv:1706.03762 [cs.CL]
Fisher, Katie. "Generate Me A Story: OpenAI’s ChatGPT Explodes in Popularity." December 16, 2022. Accessed April 18, 2024.
Chow, Andrew R. "ChatGPT Is the Fastest-Growing App Ever, and It’s Not Close." Time, February 8, 2023, 4:40 PM EST. Accessed April 18, 2024. https://time.com/6253615/chatgpt-fastest-growing/.
Zhao, Wayne Xin, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, et al. "A Survey of Large Language Models." arXiv (2023): 2303.18223.
Köbis, Nils, Christopher Starke, and Iyad Rahwan. "Artificial Intelligence as an Anti-Corruption Tool (AIACT): Potentials and Pitfalls for Top-down and Bottom-up Approaches."
Khanchel, Imen, and Naima Lassoued. 2022. "ESG Disclosure and the Cost of Capital: Is There a Ratcheting Effect over Time?" Sustainability 14, no. 15: 9237. https://doi.org/10.3390/su14159237
European Commission. "European Green Deal." Accessed April 23, 2024. https://commission.europa.eu/strategy-and-policy/priorities-2019-2024/european-green-deal_en.
Kiili, Silva. "Motivation through personal values: the Estonian Defence League, the Estonian Defence Forces and the Danish Home Guard." July 2015. https://icds.ee/wp-content/uploads/2014/Silva_Kiili_- _Motivation_through_personal_values.pdf.
Zou, Yi, Mengying Shi, Zhongjie Chen, Zhu Deng, ZongXiong Lei, Zihan Zeng, Shiming Yang, HongXiang Tong, Lei Xiao, and Wenwen Zhou. "ESGReveal: An LLM-based approach for extracting structured data from ESG reports." arXiv:2312.17264v1 [cs.CL] (25 Dec 2023).
20 Camacho, Gabriela. "Anti-corruption in ESG standards." Transparency International Anti-Corruption Helpdesk Answer, reviewed by Kaunain Rahman, Transparency International, 07 July 2022.
Voronkova, V., Oleksenko, R., & Fursìn, A. (2021). Formation of the Concept of the Socially Responsible State as a Factor of Increasing the Public Governance and Administration Efficiency. Humanities Studies, 7(84), 616-036.21-044.372. https://doi.org/10.26661/hst-2020-7-84-13
Cecchini, A., Smith, J., & Johnson, L. (2022). Title of the Article. Journal Name, Volume Number(Issue Number), Page Range
Camacho, Gabriela. "Anti-corruption in ESG standards." Transparency International Anti-Corruption Helpdesk Answer, reviewed by Kaunain Rahman, Transparency International, 07 July 2022.
Review of anti-corruption strategies Rob McCusker
McCusker, Rob. "Review of Anti-Corruption Strategies." Australian Institute of Criminology, 3 Jan. 2007.
Edmans, Alex. "The End of ESG." Financial Management 21 December 2022. https://doi.org/10.1111/fima.12413
Katz, David A., and Laura A. McIntosh. "Integrating ESG Into Corporate Culture: Not Elsewhere, but Everywhere." Wachtell, Lipton, Rosen & Katz, March
Schein, Edgar H. Organizational Culture and Leadership. 4th ed. Illustrated. The Jossey-Bass Business & Management Series. John Wiley & Sons, 2010.
Chhibber,Ajay. Governance and anti-corruption : ways to enhance the World Bank's impact (English). IEG evaluation brief Washington, D.C. : World Bank Group. http://documents.worldbank.org/curated/en/646781468330980665/Governance-and-anti-corruption-ways-toenhance-the-World-Banks-impact
McKinsey & Company. (2022, August 10). Does ESG really matter—and why? https://www.mckinsey.com/business-topics/sustainability/esg/does-esg-really-matter-and-why
Frost, K., & Spence, C. (2018). Engaging with sustainability: Stakeholder engagement in ESG reporting. Corporate Communications: An International Journal, 27(4), 241-257.
Gallup. "Organizational Culture." Gallup, January 2024, https://www.gallup.com/workplace/327371/how-tobuild-better-company-culture.aspx. Accessed [February 2024
Wong, Kellie. "Organizational Culture: Definition, Importance, and Development." Achievers, June 28, 2023, https://www.achievers.com/blog/organizational-culture-definition/. Accessed February 15, 2024.
McKelvey, Dana. "Inside Apple's New Campus: A Look at the Spaces Designed to Foster Collaboration." Fast Company, September 6, 2017, https://www.fastcompany.com/90159439/a-first-look-inside-apple-park. Accessed February 15, 2024.
Findcourses.com. U.S. L&D Report 2019: Learning Trends and Benchmarks. Accessed February 12, 2024, https://www.findcourses.com/prof-dev/download-us-ld-report-19-15209
Nirola, Narayan, and Surjit Sahu. "The interactive impact of government size and quality of institutions on economic growth- evidence from the states of India." Heliyon 5, no. 3 (2019): e01352. doi:10.1016/j.heliyon.2019.e01352.
Guan, Yanchu, Dong Wang, Zhixuan Chu, Shiyu Wang, Feiyue Ni, Ruihua Song, Longfei Li, Jinjie Gu, and Chenyi Zhuang. "Intelligent Virtual Assistants with LLM-based Process Automation." Ant Group, Renmin University of China, 2023.
"Corruption." Encyclopedia Britannica, Encyclopedia Britannica, Inc., www.britannica.com/topic/corruptionlaw.
Shleifer, Andrei, and Robert W. Vishny. "Corruption." Quarterly Journal of Economics, vol. 108, no. 3, 1993, pp. 599-617.
D. North, J. J. Wallis, and B. Weingast, Violence and Social Orders (Cambridge University Press, 2009)
Lough, John, and Vladimir Dubrovskiy. "Are Ukraine’s Anti-corruption Reforms Working?" Russia and Eurasia Programme, November 2018, Chatham House, https://www.chathamhouse.org/sites/default/files/2021- 04/2018-11-19-are-ukraine-anticorruption-reforms-working-lough.pdf
Åslund, A. (2014). The Maidan and Beyond: Oligarchs, Corruption, and European Integration. Journal of Democracy 25(3), 64-73. https://doi.org/10.1353/jod.2014.0055.
European Court of Auditors. "Reducing Grand Corruption in Ukraine: Several EU Initiatives, but Still Insufficient Results." 2021. Accessed April 29, 2024. https://www.eca.europa.eu/Lists/ECADocuments/SR21_23/SR_fight-against-grand-corruption-inUkraine_EN.pdf
Lough, John. "Ukraine’s System of Crony Capitalism: The Challenge of Dismantling ‘Systema’." Russia and Eurasia Programme, July 2021, https://www.chathamhouse.org/sites/default/files/2021-07/2021-07-01-ukrainecrony-capitalism-lough.pdf
Maksak, Hennadiy, and Sergiy Gerasymchuk. "Ukraine Disinformation Resilience Index." Foreign Policy Council "Ukrainian Prism", July 31, 2018. Accessed from: https://prismua.org/en/english-ukrainedisinformation-resilience-index/
Wolczuk, Kataryna, and Darius Žeruolis. "Rebuilding Ukraine: An Assessment of EU Assistance." Ukraine Forum, August 2018. Accessed April 29, 2024. https://www.chathamhouse.org/sites/default/files/publications/research/2018-08-16-rebuilding-ukraine-euassistance-wolczuk-zeruolis.pdf
European Court of Auditors. "Reducing grand corruption in Ukraine: several EU initiatives, but still insufficient results." Accessed on 29th of April 2024. https://www.eca.europa.eu/Lists/ECADocuments/SR21_23/SR_fight-against-grand-corruption-inUkraine_EN.pdf
"Public administration in Ukraine: Assessment against the Principles of Public Administration." December 2023.
Åberg, Martin, Ambassador of Sweden to Ukraine, Måns Nilsson, Executive Director of SEI, Ruslan Strilets, Minister of Environment Protection and Natural Resources of Ukraine, and Yevhenii Fedorenko, Deputy Minister of Environmental Protection and Natural Resources of Ukraine. "Launching the Green Agenda project for EU accession." October 23, 2023.
Corporate Governance Professional Association (CGPA) and the Centre for CSR Development Ukraine (CSR Ukraine). "ESG Transparency Index Ukraine 2020." 2020. Accessed March 1, 2024.
Cantillon, Tom, Philippe Moseley, and Janis Aizsalnieks. "Event Organized with the Contribution of United Nations Industrial Development Organization." September 2021.
Pyman, Mark. "Addressing Corruption in Military Institutions." Public Integrity 19, no. 5 (2017): 513-528. doi:10.1080/10999922.2017.1285267
General, Ryan. "Nvidia CEO worth $36 billion: 'Pain and suffering' of entrepreneurship not worth it." NextShark, October 27, 2023.
Hannah Snyder, Literature review as a research methodology: An overview and guidelines, Journal of Business Research, Volume 104, 2019, Pages 333-339, ISSN 0148-2963, https://doi.org/10.1016/j.jbusres.2019.07.039. (https://www.sciencedirect.com/science/article/pii/S0148296319304564)
Hadi, Muhammad Usman, et al. "A Survey on Large Language Models: Applications, Challenges, Limitations, and Practical Usage." 2023. https://doi.org/10.36227/techrxiv.23589741.v1.
Journal of Applied Learning & Teaching Vol. 6, no. 1 (2023). ISSN: 2591-801X. Journal of Applied Learning & Teaching. JALT. Accessed from http://journals.sfu.ca/jalt/index.php/jalt/index. Authors: Rasul, Tareq; Nair, Sumesh; Kalendra, Diane; Robin, Mulyadi; Santini, Fernando de Oliveira; Ladeira, Wagner Junior; Sun, Mingwei; Day, Ingrid; Rather, Raouf Ahmad; Heathcote, Liz.
Hern, Alex. "Techscape: AI Gadgets Need to Show Their Humane Side." The Guardian, April 16, 2024. Accessed April 30, 2024. https://www.theguardian.com/technology/2024/apr/16/techscape-ai-gadgest-humaneai-pin-chatgpt
Aspridis, George M., and Louisa Pouliana. "The Culture of Modern Public Administration: The Case of Greece." University of Thessaly, Greece, Hellenic Open University (HOU), Greece, Directorate of Secondary Education of Trikala, Greece.
O'Riordan, Joanna. "Organisational Culture and the Public Service." State of the Public Service Series, November 2015, Research Paper No. 16
Katzenbach, Jon R., Ilona Steffen, and Caroline Kronley. "Organizational Change: Cultural Change That Sticks." From the Magazine (July–August 2012).
Yang, Lihua. "Public Administration as a Dynamic Balance and Integrative Science Across Politics, Management, and Law: Rosenbloom’s Framework and Chinese Experiences." The American Review of Public Administration 49, no. 2 (March 2018): 027507401875933. doi:10.1177/0275074018759337.
Rillig, Matthias C., Marlene Ågerstrand, Mohan Bi, Kenneth A. Gould, and Uli Sauerland. "Risks and Benefits of Large Language Models for the Environment." Environmental Science & Technology 57, no. 9 (February 23, 2023): 3464–3466. https://doi.org/10.1021/acs.est.3c01106.
Carullo, G. "Large Language Models for Transparent and Intelligible AI-Assisted Public Decision-Making." 2023.
Cao, Charles, Jie Zhuang, and Qiang He. "LLM-Assisted Modeling and Simulations for Public Sector Decision-Making: Bridging Climate Data and Policy Insights." Published: 12 Dec 2023. Last Modified: 26 Feb 2024.
Snyder, Hannah. "Literature Review as a Research Methodology: An Overview and Guidelines." Journal of Business Research. Volume 104, 2019, Pages 333-339. https://doi.org/10.1016/j.jbusres.2019.07.039.
Sarnek, Anna, and Cristina Dolan. "Cybersecurity is an Environmental, Social and Governance Issue. Here's Why." March 1, 2022.