As AI systems become more deeply integrated into our high-stakes decision-making, the opacity of these models presents a growing challenge. Whether it is an autonomous vehicle navigating a storm or an adaptive tutoring system grading a student’s code, we frequently find ourselves facing a “black box” and asking: Why did the AI make this specific recommendation, and can I trust it?
In my ongoing doctoral research at SUTD, I am exploring how we can move beyond these monolithic, opaque language models. My latest work focuses on a design science framework called Explicit Orchestrated Decision Design (EODD).
The goal of EODD is to decompose complex AI decisions into specialized, inspectable roles (such as a Reasoner, Checker, and Explainer). Instead of a single, hidden output, the system generates a transparent “decision trace” that humans can actively audit, challenge, and override if necessary.
To empirically validate whether this framework actually improves human oversight, I am running a comparative online study—and I need your help.
About the Study
If you have ever been curious about how we can design AI to be more transparent and accountable, this is a great opportunity to test out some cutting-edge interface concepts.
- The Task: You will review 5 AI decision-making scenarios across two different system interfaces. Your goal is to actively spot potential AI errors (even the highly confident ones) and rate the systems on interpretability and trustworthiness.
- The Domains: You will be assigned to evaluate scenarios in either Autonomous Vehicles, AR/VR, or Learning Systems.
- Time Commitment: The online survey takes approximately 30–45 minutes to complete.
Who Should Participate?
We are currently recruiting:
- Graduate-level peers and researchers
- Teaching Assistants
- Domain novices or professionals with an interest in AI, HCI, and system design.
Your feedback will directly contribute to research aimed at making AI systems safer and more accountable to human authority.
👉 Click here to read the participant info sheet and start the survey!
Thank you for your time and for supporting human-centred AI research. Please feel free to share this link with anyone in your network who might be interested!