Governments worldwide are embracing the use of Artificial Intelligence (AI) in public services, with chatbot implementations like ChatGPT and GPT-4 showing promise but also revealing challenges in accuracy and reliability. Experts urge caution and suggest AI should complement rather than replace human civil servants to maintain accountability and reduce errors.
Governments worldwide are increasingly exploring the use of Artificial Intelligence (AI) to enhance public services. The introduction of generative AI, like OpenAI’s ChatGPT, promises more human-like interactions compared to earlier, simpler chatbots.
In the UK, the Government Digital Service (GDS) trialed a ChatGPT-based bot called GOV.UK Chat, engaging citizens in queries about benefits, taxes, and other services. Early results showed that nearly 70% of participants found the responses useful. However, there were notable issues with the chatbot generating incorrect information, raising concerns over its reliability and the potential for misplaced public confidence.
Portugal has also launched an AI-driven chatbot via the Justice Practical Guide, funded by the European Union’s Recovery and Resilience Facility (RRF). The tool, based on GPT-4, addresses basic legal inquiries, including marriage, divorce, and setting up companies. Over 28,000 questions were asked in its first 14 months, demonstrating moderate success but also occasional limitations in providing accurate answers.
Experts like Colin van Noordt and Sven Nyholm emphasize caution, suggesting these chatbots should supplement rather than replace human civil servants, due to concerns about accountability and potential for errors.
Estonia, a pioneer in digital services, employs a different approach with its Bürokratt chatbots, which use Natural Language Processing (NLP) instead of large language models. These chatbots break down user requests into key segments to infer the needed information and escalate queries to human agents when necessary.
The landscape of AI in government services is evolving, with varying implementations and ongoing efforts to balance efficiency with reliability.