Towards Robust identification and Unraveling of human-System Trust correlates (TRUST)

Description

The increasing technology development during the last decades brings us to interact daily with automated systems. These systems are becoming more and more reliable, relegating operators to the role of passive supervisors and excluding them from the control loop. Moreover, their increasing complexity has constrained the designers of these systems to make them more and more obscure, imposing on the operators a blind trust in the decisions made. However, trust in automated systems poses major problems because it is regularly subject to calibration errors. Numerous research studies, both fundamental and in ecological environments, have demonstrated that this over-trust in highly reliable systems can be the cause of operational difficulties such as the inability to detect (infrequent) system errors when they appear. Similarly, under-trust, or even mistrust, can also be deleterious and reduce the operational performance of the operator-automated system couple. However, the mechanisms supporting the emergence of trust are still unclear and little studied. In particular, the neurophysiological markers associated with trust are poorly identified and validated. The literature shows that research on these correlates has not focused on characterizing trust, but on identifying its impact on specific cognitive processes, such as error detection or the feeling of control.
One of the difficulties related to this characterization comes from the field of Brain-Computer Interfaces (BCI), where promising advances have been made in the last ten years, especially concerning passive BCIs which remain a major challenge in Machine Learning (ML) based on brain activity measurements.
In this thesis, we seek to identify the brain correlates of trust in highly automated reliable systems, as well as their variations over time, based in particular on measurements of brain activity (electroencephalography, or EEG). A major emphasis is put on the characterization, understanding and evaluation of trust in a transversal way independently of the type of task or cognitive process involved. We therefore wish to determine brain correlates of trust, in the general sense of the term, and which could be measured in real-time on operators.

Nature de finanement

Doctoral school

Desired profile

Candidates should have a Master’s degree in Cognitive Science, or a Master’s degree in Machine Learning (or equivalent) with experience in Signal Processing and Statistics. Programming skills (Python or MATLAB) are required. Experience in electroencephalographic (EEG) data acquisition would be appreciated but is not mandatory.

Company

ISAE-SUPAERO is a research and innovation-driven institution of higher education, committed to balancing scientific excellence, academic visibility and proximity to industrial objectives. The Department of Aerospace Vehicles Design and Control (DCAS) is developing training and research activities to meet the scientific challenges on tomorrow’s air transportand future space systems. The department integrates environmental and social-economic dimensions into its studies such as reducing environmental impacts (consumption, noise, emissions), optimizing the design cycle and operating costs, increasing the safety of aerospace systems and improving the efficiency of human-machine systems.The DCAS includes 35 permanent staff members (faculty, researchers, and engineers, technical and administrative staff) and around 45 non-permanent members including engineers on a contract basis, doctoral and post-doctoral students.
The DCAS manages a fleet of nine aircraft (1 Socata TB-20, 3 Robin DR- 400, 4 Aquila, 1 twin- engine Vulcan Air P68 Observer) at the Lasbordes Operational Center and develops numerous simulators and software for the training and research needs of the department.