My background is in pure math and I’m excited to combine it with philosophy and psychology in my PhD. In particular, I am interested in psychology-inspired AI. Insight into Theory of Mind, decision-making, reasoning, learning and language understanding in humans can be used to inform the design and build of AI’s with similar skills.
In the past year I have explored the following problems:
– Designing an argumentation framework that captures the role that arguments play in our daily life, which is updating our beliefs.
– Finding a way to synergistically combine two value learning methods: inverse reinforcement learning based on human demonstrations, and collecting human judgements of debates between expert AI systems.
– Developing a mechanism through which to construct symbolic arguments (for use by AI debaters), by abstracting from sub-symbolic data.
– Designing environments and reward functions in which RL agents are encouraged to learn how to use language, reason and argue (a field known as emergent communication). I am mostly method-agnostic and have so far worked with deep learning, reinforcement learning, agent-based modelling, category theory, game theory and argumentation theory.
