How Can AI Improve Government Transparency?

Corruption remains one of the most serious threats to the development and stability of any country. It undermines trust in public institutions, leads to inefficient use of resources, and hinders economic development. 

Traditional methods in the fight against corruption often prove insufficient due to the complexity and scale of the problem. In this context, artificial intelligence (AI) becomes an indispensable tool, offering new possibilities for identifying and preventing corruption risks. Through powerful data analysis algorithms, machine learning (ML), and process automation, AI can significantly enhance the effectiveness of anti-corruption measures, ensuring transparency, accountability, and a quick response to potential violations. In this article, we will explore why the use of AI is critically important in the modern fight against corruption and how exactly it can assist in this process.

Research shows that AI can be useful both as a preventive tool in the fight against corruption and as a tool for detecting misconduct. As a preventive tool, AI can accurately predict patterns of corruption. For example, scientists have used AI to develop an early warning system to predict corruption using political and economic factors such as economic growth and the stability of a political party [1].

Recent studies have also shown that ML can successfully predict conflicts of interest in public procurement [2] and the formation of criminal groups in the public procurement sector [3]. 

As a tool for detecting misconduct, AI can identify anomalies that are manifestations of corruption and related phenomena, particularly in public procurement [4]. AI technologies can detect fake suppliers [5], and ML models can successfully identify collusion between private companies and government agencies [6], as well as other signs of corruption.

AI offers numerous advantages to the public sector for ensuring organizational or individual compliance with established regulatory requirements, standards, and internal policies. With AI, it is possible to improve the oversight and evaluation process in government institutions regarding adherence to ethical standards, transparency, and compliance with laws and other regulations.

AI for public services includes methods such as ML, deep learning, computer vision, speech recognition, natural language processing (NLP), and robotics. The primary reason for using AI in monitoring government processes is that it can free up millions of work hours. This can allow public servants to focus on more important tasks and lead to faster service delivery to the population by the government.

According to estimates by Deloitte [7], automating the tasks of public servants could save approximately 96.7 million to 1.2 billion work hours, potentially leading to annual savings of around 3.3 to 41.1 billion U.S. dollars.

Artificial intelligence can significantly contribute to ensuring government transparency through several key methods:

1. Automation and Data Analysis

AI can process large volumes of information much faster and more accurately than humans. This capability allows government agencies to automate routine processes such as data collection, storage, and analysis. For instance, AI systems can automatically process financial reports, budget data, and documents, identifying potential anomalies and discrepancies. This greatly simplifies the audit process and allows for quicker detection of corrupt schemes.

Examples  of using AI for automation and data analysis

1. The U.S. Internal Revenue Service (IRS) uses AI to speed up and assist with routine verification processes that are too complex or extensive for the agency’s current capabilities. AI’s capabilities include analyzing complex financial data, tax returns, and offshore reports. The use of AI has enabled the IRS to determine that 75% of taxpayers involved in digital asset exchanges do not comply with federal tax laws.

2. AGESIC, Uruguay’s agency for electronic government and information and public knowledge, aimed to digitize all public services. This initiative involved automating the classification of support service requests using AI methods such as ML and NLP. The integration of AI technologies allowed for the automation of administrative tasks, document verification, and data entry, thereby freeing up government resources to focus on more complex and value-added activities. This shift towards automation led to increased efficiency, reduced operational costs, and accelerated service delivery, ultimately enhancing AGESIC’s overall productivity.

3. National Center for Missing & Exploited Children (NCMEC) uses an AI tool called Logikcull, developed by Reveal, to sift through large volumes of data and identify the necessary information for criminal investigations. This tool has significantly reduced the time and resources required to identify and prosecute child abductors on a large scale.

4. The Spanish National Police implemented a tool called VeriPol in police stations to help detect false complaints, such as individuals falsely reporting robberies. VeriPol uses natural language processing, a type of ML technology that helps AI systems understand and interpret human language. The algorithms analyze this language based on historical police reports entered into the system. As a result of implementing this tool, there has been a significant increase in the detection of false reports and time savings for the police. The tool identifies false robbery reports with an accuracy of over 80 percent.

5.  The Trelleborg Social Welfare Department in Sweden used AI technologies, including robotic process automation (RPA), to automate various social assistance decisions. This innovation significantly reduced the waiting time for aid applications, providing more efficient citizen services and reducing the burden on staff.

2. Predicting and Preventing Corruption Risks

With the help of ML algorithms, AI tools can predict potential corruption risks based on historical data and behavioral patterns. This capability allows government agencies to take proactive measures to prevent violations, rather than merely responding after the fact. For example, AI systems can analyze procurement processes, detecting suspicious connections between participants or unusual changes in contract values.

Example of using AI for predicting and preventing corruption risks

In Ukraine, the online platform DOZORRO uses AI to monitor procurement data, identifying signs of corruption and irregularities in the procurement process. By quickly scanning large volumes of data, the system significantly enhances the efficiency of expert analysis of tenders. The automation of monitoring and data analysis processes has led to a 26% increase in detected tenders with unjustified winner selection, a 37% increase in those with unwarranted disqualifications, and a 298% increase in cases of collusion among participants.

3. Enhancing Transparency through Open Data

AI is also effectively used in supporting open data platforms where citizens can access information about government activities. Such platforms can leverage AI technologies to organize and analyze large datasets, making the information more understandable and accessible to the general public. This includes data visualization and providing user-friendly interfaces. 

By utilizing AI in open data platforms, governments can make vast amounts of information more digestible for citizens, fostering greater public engagement and scrutiny, which in turn can act as a deterrent to corrupt practices.

Examples  of using AI for enhancing transparency through open data

1. The Belgian company CitizenLab has created a public engagement platform based on ML algorithms that helps public officials effectively manage and utilize thousands of citizen requests for decision-making. CitizenLab developed its own natural language processing (NLP) methods, using ML and AI to quickly extract key ideas from large volumes of unstructured citizen input, such as ideas, comments, and votes. This AI-driven digital participation platform enables governments to make more informed decisions while saving time and resources.
2. The Citibeats AI platform analyzes data to identify social facts and trends valuable to companies and institutions. The platform interprets user needs and opinions, adapting to the context in which they are expressed, considering linguistic requirements, data sources, and text structure. The Citibeats algorithm combines NLP and ML technologies to filter relevant content, classify user opinions and information, and automatically extract ideas and patterns. This technology ensures a more accurate detection of the real impact of any decision or action taken.

4. Supporting Citizen Engagement

Interactive AI-based chatbots and virtual assistants can enhance interactions between government agencies and citizens. They can provide quick responses to inquiries, help find necessary information, and ensure feedback. This improves service levels and ensures a more transparent and efficient communication process.

Examples  of using AI to support citizen engagement

1. The Revenue Commissioners of Ireland used an AI-based voice bot to provide flexible on-demand customer service. Integrated with the company’s telephone system and back-end transaction system, this AI voice bot ensures immediate availability and real-time updates on customer inquiries.

2. The AI-based language technology platform of the Latvian State Administration provides round-the-clock support to thousands of clients, offering advice and information. This platform offers automatic translation, speech recognition and synthesis, as well as various multilingual support tools for e-services.

5. Comprehensive Infrastructure Monitoring

Governments worldwide face challenges in tracking real estate. Manual management is complex and often insufficient for detecting land developments. These challenges present opportunities for AI implementation in government systems to automate property monitoring and management.

Examples  of using AI for comprehensive infrastructure monitoring

1. The Southeast Michigan Council of Governments (SEMCOG) used Nearmap and Ecopia AI to identify unregistered buildings in their region. Using high-resolution aerial imagery and AI-based mapping technology, Nearmap helped SEMCOG integrate building outlines with existing geospatial data. AI-driven analysis enabled the government to identify over 3,452 previously unregistered buildings in the region.

2. In France, an AI tool utilizing satellite mapping was applied to detect tax evaders. French tax authorities discovered 20,356 undeclared private swimming pools, allowing them to collect approximately 10 million euros in taxes.

Conclusion

AI opens new possibilities for improving the transparency and accountability of government agencies. AI enables process automation, enhances data analysis, predicts risks, and provides open access to information for citizens. This not only increases government efficiency but also strengthens public trust in state institutions, which is a key factor in the development of a democratic society. The implementation of AI in public administration is an important step towards a more transparent, accountable, and effective government.

Sources:

^[1] Félix J. López-Iturriaga & Iván Pastor Sanz (2018). “Predicting Public Corruption with Neural Networks: An Analysis of Spanish Provinces,” Social Indicators Research: An International and Interdisciplinary Journal for Quality-of-Life Measurement, Springer, vol. 140(3), pages 975-998, December. https://ideas.repec.org/a/spr/soinre/v140y2018i3d10.1007_s11205-017-1802-2.html

^[2] Mazrekaj, Deni and Titl, Vítězslav and Schiltz, Fritz, Identifying Politically Connected Firms: A Machine Learning Approach (June 4, 2021). https://ssrn.com/abstract=3860029 or http://dx.doi.org/10.2139/ssrn.3860029

^[3] Mihály Fazekas 1 , Bence Tóth 2 , Johannes Wachs (2023). Public procurement cartels: A large-sample testing of screens using machine learning. April 2023, Budapest, Hungary. https://www.govtransparency.eu/wp-content/uploads/2023/04/Fazekas-et-al_PP-cartel-detection_GTI-WP_2023.pdf

^[4] Kossow, Niklas. “Digital anti-corruption: hopes and challenges.” In A Research Agenda for Studies of Corruption by Alina Mungiu-Pippidi and Paul Heywood (eds.). Cheltanham, UK: Edward Elgar Publishing, 2020. https://doi.org/10.4337/9781789905007.00019

^[5] Dominik Olszewski (2014). Fraud detection using self-organizing map visualizing the user profiles. Knowledge-Based Systems. Volume 70, November 2014, Pages 324-334. https://www.sciencedirect.com/science/article/abs/pii/S0950705114002652

^[6] Manuel J. García Rodríguez a, Vicente Rodríguez-Montequín a, Pablo Ballesteros-Pérez b, Peter E.D. Love c, Regis Signor (2022). Collusion detection in public procurement auctions with machine learning algorithms. Automation in Construction. Volume 133, January 2022, 104047. https://www.sciencedirect.com/science/article/pii/S0926580521004982

 ^[7]  William D. Eggers, David Schatsky, Peter Viechnicki. AI-augmented government. https://www2.deloitte.com/us/en/insights/focus/cognitive-technologies/artificial-intelligence-government.html