BlueBird
BlueBird is a project that aims to develop an AI system for air traffic control. Crucial to this aim is the design and development of methods and tools that promote safe, explainable and trustworthy use of AI in this domain. Bluebird is £13.75 project led by Prof. Tim Dodwell, and it is supported through an investment from UK EPSRC in a partnership between NATS and The Alan Turing Institute.
Moral Machine
Moral Machine is an online platform for gathering a human perspective on moral decisions made by autonomous vehicles (AVs). It presents users with moral dilemmas that are faced by AVs. The dilemmas are inspired by the Trolley Problem, a famous philosophical conundrum. Since its deployment in June 2016, the website has received high publicity and coverage. This has produced the largest dataset on ethics of machines ever collected: 100 million decisions in ten languages from millions of people in 233 countries and territories.
MyGoodness
MyGoodness is an online platform created in collaboration with Peter Singer's foundation, The Life You Can Save. MyGoodness proposes dilemmas in a similar fashion to Moral Machine in the form of possible charities. The goal of MyGoodness is to quantify the effect of the different factors that may influence people to give ineffectively, and to raise awareness among the public about how people can be very ineffective when giving. Since its deployment in December 2017, MyGoodness has been visited by 250,000 users who have contributed over three million responses.
Blaming Humans and Machines
When something wrong happens as a result of a decision, how are blame and responsibility attributed to humans and machines involved in making the decision? The answer depends on various factors including the domain, whether they share the task, and the difficulty of the task. My work considers a variety of hypothetical situations characterised by these factors.
Opinion Aggregation
Various mechanisms have been implemented to capture individuals' opinions online such as thumbs-up/down. However, in more contested domains (e.g., Wikipedia, and political discussion) these mechanisms are not sufficient since they only deal with each issue independently. In such cases, the use of arguments provides a better tool to represent the conflict on the Web. I investigated the theoretical limits of such a type of aggregation, and how to circumvent these limits.