Courses & seminarsStudentship
ESRC Studentship to investigate user trust in artificial intelligence systems

UPDATE -: We are no longer taking accepting expresssions of interest for this post.

A fantastic opportunity to work with a great team on an issue of growing public significance

  • A four year ESRC studentship opportunity
  • Can we / should we / how should we - trust what the machine is telling us?
  • Linked to a large Wellcome Trust funded research project which is building an AI system in behaviour change
  • Supportive environment to develop your own research interests

The proposed studentship will be embedded within a major interdisciplinary research project funded by Wellcome Trust: the Human Behaviour-Change Project (HBCP). This is a collaboration between computer science, behavioural science and system architecture. The student will be based at the EPPI-Centre, UCL, which has a long-standing stream of work on information science in evidence synthesis. The HBCP is developing a free, online system that people can use to query the global evidence on behaviour change. The system aims to offer a major step forward in the accessibility of research evidence to decision-makers, and also offer researchers the ability to identify all research about a given behaviour change intervention far more quickly than is currently possible. This is clearly an ambitious aim, which will depend on Artificial Intelligence (‘AI’) tools being developed at IBM Research Dublin for its success. Its success is also dependent on users’ perceptions of the system and, critically, the degree to which they trust its outputs.

We have put together a possible structure for this studentship (below), but would also stress that we would expect the student to take ownership of their own research and that they may choose to follow a different structure. Although this studentship is clearly grounded within advances in computer science, its focus means that candidates could specialise within the realm of the sociology/psychology of new technologies, organisational psychology/sociology, or the psychology/sociology of decision-making. As long as the overall research is of high standard and within the remit of the overarching aims of the studentship, we will encourage and support the appointed student in shaping the PhD around their own interests. For example, the student may want to spend some time developing technical AI skills of their own. We have sufficient expertise, data and source code within the team to support this, if the student chooses to follow this route.

Rather than building the AI system itself, this studentship will centre on the social scientific aspects of the project and in particular on the issue of trust. Deployments of similar systems to the one being developed – where users are able to access recommendations for action based on far more data than they would be able to synthesise themselves – have shown that users tend to trust recommendations that correspond with their own prior judgements (possible confirmation bias); but in situations where the system’s recommendations are at variance with their own, will tend then to distrust the reliability of the system, rather than reassess their own assumptions. The latter may be perfectly rational and appropriate, but does call into question the value that such systems can add in these situations. The student will carry out in-depth data collection and analysis with researchers and decision-makers to explore these issues of trust and warrant, addressing two main sets of research questions:

  1. How do users interact with an AI system which contains recommendations for action? (Do they trust what it says? What kind of supporting evidence does the system need to supply to engender trust? What type of actions do users feel comfortable making based on the system’s recommendations? Are there areas where users will not trust the system at all? Can recommendations from the system ever change a user’s prior assumptions? (if so, when and how?)
  2. How can the AI system ‘explain’ itself to users in a way that correctly communicates the degree of uncertainty within the system associated with specific recommendations?  Does this increase trust?

Addressing these questions will involve joint work between the student and IBM Research Dublin and frequent interaction and engagement with end users of the system. The student will have opportunities to work with decision-makers at the interface between cutting-edge computer science and how AI outputs are perceived and used. The student will also have the opportunity for social and behavioural input from and collaboration with behavioural and social scientists in the HBCP. Significant knowledge exchange is planned, including spending time at the IBM Research site in Dublin.

We have put together a possible structure for this studentship, but would also stress that we would expect the student to take ownership of their own research and that they may choose to follow a different structure. Although this studentship is clearly grounded within advances in computer science, its focus means that candidates could specialise within the realm of the sociology/psychology of new technologies, organisational psychology/sociology, or the psychology/sociology of decision-making. As long as the overall research is of high standard and within the remit of the overarching aims of the studentship, we will encourage and support the appointed student in shaping the PhD around their own interests. For example, the student may want to spend some time developing technical AI skills of their own. We have sufficient expertise, data and source code within the team to support this, if the student chooses to follow this route.
 

Information on eligibility can be found here: https://esrc.ukri.org/skills-and-careers/studentships/prospective-students/am-i-eligible-for-an-esrc-studentship/

This is a “1+3” studentship, meaning that, if you don’t have an ESRC recognised research training qualification (or equivalent), you can undertake research methods training during the first year of a four-year award.

The studentship will start in October 2018 (soon!), or as soon as possible afterwards. We are currently finalising our recruitment process, but October is fast approaching, so if you are interested in applying, do get in touch as soon as possible with one of the supervisors: James Thomas (james.thomas@ucl.ac.uk), Susan Michie (s.michie@ucl.ac.uk) or Dylan Kneale (d.kneale@ucl.ac.uk).

UPDATE: We are no longer accepting expressions of interest for this post

Copyright 2016 Social Science Research Unit, UCL Institute of Education :: Privacy Statement :: Terms Of Use :: Site Map :: Login
About::Projects::Publications::Research Use::Courses & seminars::Resources::Databases::Blog