Making Moral Machines Transparent

Artificial Intelligent Agents (AIAs) are being developed which make decisions of moral salience. These decisions have increasingly larger impact on our lives at an individual, societal and global level. Yet, we often hold inaccurate mental models of such agents. In this current research project, supervised by Andreas Theodorou, I am exploring how our perceptions of a decision-maker are impacted by factors such as agent type (Natural vs. Human) and transparency implementation (whether the agent's moral framework is divulged).
To do so, we have created a virtual reality self-driving car simulation, in which the user experiences multiple moral dilemmas. This software will eventually be made open source for other researchers.
We conducted experiments at the Fantastical Multimedia Pop-Up Project, which were featured in this news article.


fMRI Investigation: Blind in the Mind's Eye

Close your eyes. Imagine a lion. Concentrate on the rich colours of its fur and low rumble of its purr. How vividly could you form this mental image?... I investigated the neural correlates of the syndrome Aphantasia. This syndrome is characterised by a deficit in visual imagination. By using functional MRI (a brain imaging technique) on participants with and without aphantasia I was able to gain insight into the syndrome. Brain imaging was conducted whilst they visually imagined and perceived images of famous places and famous faces. I then preprocessed, analysed the data using SPM8 and SPSS software, then interpreted the results. This research was part of the wider project 'The Eye's Mind' ongoing at Exeter Medical School. See the poster for a more indepth summary.