By T. Jayani, JadeTimes News
Long before the advent of ChatGPT, governments were eager to employ chatbots for automating their services and providing advice. These early chatbots were relatively simple, with limited conversational abilities, according to Colin van Noordt, a researcher specializing in the use of AI in government based in the Netherlands.
The emergence of generative AI in the past two years has revived the vision of more efficient public service, featuring human like advisors available around the clock to answer questions about benefits, taxes, and other government related interactions. Generative AI's sophistication allows it to provide human like responses and, if trained on sufficient quality data, it could theoretically address a wide range of queries regarding government services.
However, generative AI is also known for making mistakes or providing nonsensical responses, known as hallucinations. In the UK, the Government Digital Service (GDS) tested a ChatGPT based chatbot called GOV.UK Chat, designed to answer citizens' questions on various government services. According to a blog post about the early findings, nearly 70% of trial participants found the responses useful. Yet, there were instances where the system generated incorrect information presented as fact. The blog expressed concern about misplaced confidence in a system that could occasionally be wrong, emphasizing that the answers did not meet the high level of accuracy required for a site like GOV.UK, where factual correctness is crucial. The agency is rapidly iterating on this experiment to improve accuracy and reliability.
Other countries are also experimenting with generative AI systems. In 2023, Portugal launched the Justice Practical Guide, a chatbot designed to answer basic questions on topics such as marriage and divorce. Funded by the European Union’s Recovery and Resilience Facility (RRF), the €1.3 million ($1.4 million; £1.1 million) project is based on OpenAI’s GPT-4.0 language model. Besides marriage and divorce, it also provides information on setting up a company. According to the Portuguese Ministry of Justice, 28,608 questions were posed through the guide in the first 14 months. When asked basic questions like "How can I set up a company," it performed well. However, for more complex queries like "Can I set up a company if I am younger than 18 but married?" it could not provide an answer. A ministry source acknowledged the chatbot's limitations in trustworthiness, despite rare instances of incorrect replies, and expressed hope that these limitations would be overcome by increasing the answers' level of confidence.
Given these flaws, many experts advise caution. Colin van Noordt suggests that chatbots should be viewed as an additional service for quick information retrieval rather than a replacement for human workers to reduce costs. Sven Nyholm, a professor of the ethics of artificial intelligence at Munich’s Ludwig Maximilians University, highlights the problem of accountability, noting that a chatbot cannot replace a civil servant who can be held accountable and morally responsible for their actions. Public administration requires accountability, which necessitates human involvement. Nyholm also points out the reliability issue, stating that newer chatbots create an illusion of intelligence and creativity that older chatbots did not, but they can still make silly and potentially dangerous mistakes if relied upon too heavily.
