The firstFirst chatbots “tended to be simpler, with limited conversational abilities,” according to Dutch researcher Colin van Noordt, who specializes in the application of AI in government.
However, the last two years have seen the rise of generative AI, which has resurrected the idea of more effective public service—as human-like advisors available around-the-clock to address inquiries about taxes, benefits, and other matters pertaining to public-government interactions.
In theory, generative AI could address a wide range of inquiries regarding government services if it were taught on sufficient high-quality data. It is intelligent enough to provide answers that are comparable to those of a human.
However, generative AI is already widely recognized for producing errors or even absurd responses—a phenomenon known as hallucinations.
A ChatGPT-based chatbot named GOV.The Government Digital Service (GDS) in the United Kingdom has tested British Chat. It is designed to respond to inquiries from the public on a variety of topics pertaining to government services.
The agency stated that about 70% of trial participants found the comments helpful in a blog post discussing their preliminary results.
But there were issues with “a few” instances where the system produced false information and presented it as true.
The blog also expressed worry that people would have misplaced faith in a system that isn’t always accurate.
