Interview with Amar Halilovic: Explainable AI for robotics

On this interview collection, we’re assembly a few of the AAAI/SIGAI Doctoral Consortium individuals to search out out extra about their analysis. The Doctoral Consortium supplies a chance for a bunch of PhD college students to debate and discover their analysis pursuits and profession aims in an interdisciplinary workshop along with a panel of established researchers. On this newest interview, we hear from Amar Halilovic, a PhD pupil at Ulm College.

Inform us a bit about your PhD – the place are you learning, and what’s the matter of your analysis?

I’m presently a PhD pupil at Ulm College in Germany, the place I deal with explainable AI for robotics. My analysis investigates how robots can generate explanations of their actions in a method that aligns with human preferences and expectations, notably in navigation duties.

Might you give us an outline of the analysis you’ve carried out to this point throughout your PhD?

To this point, I’ve developed a framework for environmental explanations of robotic actions and selections, particularly when issues go fallacious. I’ve explored black-box and generative approaches for the technology of textual and visible explanations. Moreover, I’ve been engaged on planning of various rationalization attributes, akin to timing, illustration, length, and so forth. Recently, I’ve been engaged on strategies for dynamically selecting the right rationalization technique relying on the context and person preferences.

Is there a side of your analysis that has been notably attention-grabbing?

Sure, I discover it fascinating how individuals interpret robotic habits otherwise relying on the urgency or failure context. It’s been particularly rewarding to check how rationalization expectations shift in numerous conditions and the way we are able to tailor rationalization timing and content material accordingly.

What are your plans for constructing in your analysis to this point through the PhD – what facets will you be investigating subsequent?

Subsequent, I’ll be extending the framework to include real-time adaptation, enabling robots to study from person suggestions and alter their explanations on the fly. I’m additionally planning extra person research to validate the effectiveness of those explanations in real-world human-robot interplay settings.

Amar along with his poster on the AAAI/SIGAI Doctoral Consortium at AAAI 2025.

What made you wish to research AI, and, specifically, explainable robotic navigation?

I’ve all the time been within the intersection of people and machines. Throughout my research, I spotted that making AI methods comprehensible isn’t only a technical problem—it’s key to belief and value. Robotic navigation struck me as a very compelling space as a result of selections are spatial and visible, making explanations each difficult and impactful.

What recommendation would you give to somebody considering of doing a PhD within the subject?

Choose a subject that genuinely excites you—you’ll be residing with it for a number of years! Additionally, construct a assist community of mentors and friends. It’s simple to get misplaced within the technical work, however collaboration and suggestions are very important.

Might you inform us an attention-grabbing (non-AI associated) reality about you?

I’ve lived and studied in 4 totally different nations.

About Amar

Amar is a PhD pupil on the Institute of Synthetic Intelligence of Ulm College in Germany. His analysis focuses on Explainable Synthetic Intelligence (XAI) in Human-Robotic Interplay (HRI), notably how robots can generate context-sensitive explanations for navigation selections. He combines symbolic planning and machine studying to construct explainable robotic methods that adapt to human preferences and totally different contexts. Earlier than beginning his PhD, he studied Electrical Engineering on the College of Sarajevo in Sarajevo, Bosnia and Herzegovina, and Laptop Science at Mälardalen College in Västerås, Sweden. Exterior academia, Amar enjoys travelling, images, and exploring connections between know-how and society.