Large Language Models in Human–Computer Interaction

Personas are representative of real groups and are used to enhance empathy among decision-makers. Personas have been used for commercial and social good uses. However, persona generation has certain challenges when representing minority views and, on a larger scale, when modeling complex interactions in an ecosystem comprising multiple parties. To address these challenges, this chapter extends the use of large language models (LLMs) in persona generation in two different directions. First, LLM-generated personas are used to represent marginalized communities by providing an example of the Global South. Second, personas are expanded from the end-user perspective to all key parties in the ecosystem. This view involves the potential to generate personas for not only human entities but also for non-human entities. By modeling the ecosystem as a “playground,” personas provide an opportunity for the parties to interact with other ecosystem parties for enhanced perspective-taking and empathy propagation. When the concept of personas is extended beyond the human user, it perpetuates empathy and reduces self-referential bias, from the minority view to the scale of all the parties included in the entire ecosystem.