Is artificial intelligence the solution to improved government services.
Efficiency, accessibility, and responsiveness can all be improved by integrating artificial intelligence (AI) into public services, as has long been claimed. Modern AI has opened up new opportunities for governments around the world, especially with the introduction of Large Language Models (LLMs) like Google’s Gemini and ChatGPT. The models in question have conversational capabilities akin to those of a human being, with the ability to address a wide range of inquiries pertaining to public information, legal advice, benefits, taxes, and other government services.
However, the journey towards implementing AI in government is fraught with challenges and ethical considerations. Colin van Noordt, a researcher specializing in AI applications in government based in the Netherlands, points out the historical evolution of chatbots. Early iterations were simpler, with limited conversational abilities. The emergence of generative AI in the past two years, powered by LLMs, represents a leap forward in capability. These models, trained on vast datasets, can theoretically handle a wide array of queries, offering responses that mimic human speech patterns and understanding.
Yet, the allure of generative AI is tempered by significant drawbacks. One of the primary concerns is accuracy and reliability. Despite their sophisticated training, LLMs are prone to errors and inconsistencies, sometimes generating nonsensical responses or misinformation. In the United Kingdom, for instance, the Government Digital Service (GDS) conducted trials with ChatGPT-based chatbots under the GOV.UK Chat initiative. While generally useful, the system occasionally provided incorrect information, highlighting the challenge of ensuring factual accuracy in critical government interactions.
Sven Nyholm, a professor of AI ethics at Munich’s Ludwig Maximilians University, raises profound ethical questions regarding the deployment of AI in public administration. Unlike human civil servants who can be held accountable for their actions, AI chatbots lack moral agency and cannot bear responsibility for errors or decisions that may adversely affect citizens. This accountability gap is significant in contexts where transparency, fairness, and trust are paramount, such as legal advice or sensitive government services.
Moreover, Nyholm underscores the limitations of AI in understanding context and nuance. While LLMs can simulate intelligence and creativity to some extent, they often struggle with complex queries that require deep comprehension or subjective judgment. This limitation poses a challenge in scenarios where precise and contextually sensitive responses are essential.
Despite these challenges, governments continue to explore the potential benefits of AI-driven services. China, for instance, has made significant strides in adopting generative AI technologies. According to a survey by U.S. AI and analytics software company SAS and Coleman Parkes Research, 83% of Chinese respondents reported using generative AI in various industries, surpassing global averages significantly. This rapid adoption underscores China’s commitment to leveraging AI for technological leadership and economic competitiveness.
With its Bürokratt chatbots, Estonia presents an interesting alternate strategy to other countries that have adopted LLMs extensively. Since the early 1990s, Estonia has led the way in the development of digital government services, being a pioneer in the creation of complete e-government platforms and digital IDs. With a concentration on Natural Language Processing (NLP) as opposed to LLMs, Bürokratt is a representation of Estonia’s customized approach to AI in public services.
NLP operates differently from LLMs by breaking down user queries into structured components, identifying key words, and inferring intent based on predefined rules and datasets. This approach prioritizes reliability and accuracy, emphasizing factual information over conversational flair. Kai Kallas, head of the Personal Services Department at Estonia’s Information System Authority, explains Bürokratt’s operational model. When faced with queries beyond its capabilities, Bürokratt seamlessly transfers the interaction to human customer support agents, ensuring that users receive accurate responses without compromising on quality.
The Bürokratt project underscores Estonia’s commitment to transparency and user trust. Unlike LLMs, which often operate as black-box systems with opaque decision-making processes, NLP-based chatbots offer greater visibility into their operation. This transparency is crucial in public administration, where accountability and reliability are essential for fostering public trust in digital government services.
Nevertheless, NLP-based systems like Bürokratt are not without limitations. While they excel in providing structured and reliable information, they may lack the conversational depth and creativity of LLMs. Estonia acknowledges this trade-off but prioritizes accuracy and user confidence in its approach to AI-driven public services.
Looking ahead, the global landscape of AI in government services remains dynamic and complex. Governments must navigate technological advancements while addressing ethical concerns and ensuring equitable access to AI-driven services. The debate over the role of AI chatbots versus human interaction continues to evolve, with proponents advocating for AI’s potential to streamline operations and improve service delivery, while skeptics emphasize the irreplaceable value of human judgment and accountability.
Even though ChatGPT and other LLMs have great opportunities to improve government services through AI, user trust, accuracy, and accountability must all be carefully considered when implementing them. An alternate use of NLP that emphasizes dependability and openness in the provision of digital services is demonstrated by Bürokratt in Estonia. Realizing the full potential of AI in public administration would need governments around the world to find a balance between ethical governance and innovation in AI technologies.
If you like the article please follow on THE UBJ.
