英国成人のLLM利用、信頼、便益とリスクへの意識

Increasing numbers of people now routinely interact with large language models (LLMs) across many aspects of life, including in the workplace, educational settings, and for personal activities. The pace at which these tools have been adopted across society in recent years has led to substantial shifts in the ways in which people approach tasks and access information and advice. As a result, it is important for researchers and policymakers to gain up-to-date evidence on how the public engages with these technologies, including what people typically use them for, the extent to which users trust the information and advice that LLMs provide, and how the public perceives the potential benefits and risks associated with their use. We surveyed a nationally representative sample of 2,002 adults in the UK. Participants were asked about their use of LLMs, including both practical applications and more personal and social forms of engagement, their trust in the information provided by these systems across a range of topics and compared with other common sources, and their attitudes towards the potential societal benefits and risks associated with these tools. Results show that while practical tasks remain the most common type of LLM use, many people now engage with these tools for personal support. Almost one third of regular users (31%) say they use LLMs for personal and emotional support, such as talking through problems and asking for help with decisions, while one quarter report interacting with LLMs for meaningful conversation. We also find that trust in LLM-generated information is relatively high, but that public attitudes towards LLMs are characterised by both optimism and concern. While over three quarters of our sample report feeling enthusiastic about the potential benefits of LLMs (77%), a majority also express concern about their potential risks (70%).