Innovating at the Intersection of Emotions and AI-based Language Models

Canadians have experienced tremendous mental health challenges over the last few years. At the same time, systems to address mental health are over-burdened. There is thus a clear need to scalable and low-cost technologies that could support the mental and emotional wellbeing of Canadians. This project is about developing innovations in artificial intelligence that could support the mental and emotional wellbeing of Canadians. Artificial intelligence (AI) is making stunning advances. Germane to this proposal are advances in AI for natural language processing. More specifically, language models which can generate text have been increasing in sophistication and have been used as conversational chatbots. But these “conversational AIs” have some limitations. For example, language models stop short of actually “understanding” language. Instead, they use of probabilistic estimation of word sequences to generate text. Language models also seem to reproduce the some of the social biases that may be present in their training datasets. As these language models sometimes produce surprising and inappropriate texts, their use as agents for supporting mental health and emotional wellbeing has, to date, been limited.

Faculty Supervisor:

Piper Jackson

Student:

Partner:

Approach Analytics Inc

Discipline:

Computer science

Sector:

Information and cultural industries; Professional, scientific and technical services

University:

Thompson Rivers University

Program: