The two DTU Compute professors Aasa Feragen and Thomas Bolander are among 9 DTU researchers who have just received funding from the Independent Research Fund Denmark (Danmarks Frie Forskningsfond) for their promising and original research ideas.
With the grant, Aasa Feragen from the research section Visual Computing and Thomas Bolander from the research section for Algorithms, Logic and Graphs will have the opportunity to pursue their most innovative ideas and promote innovative Danish research.
Aasa Feragen's project: EASE: Explainable AI as a clinical mentor for obstetric ultrasound
Medical artificial intelligence (AI) often faces challenges when integrating into clinical practice. One significant reason is that AI systems are frequently developed without adequately addressing the actual clinical needs: Is artificial intelligence truly beneficial in the daily clinical workflow? In this project, clinicians collaborate with researchers in artificial intelligence and medical education to create decision support based on explainable AI.
The goal is to optimize AI explanations to enhance clinician learning - a departure from the conventional approach, which typically aims to replace clinicians.
To achieve this, we need robust methods for validating whether the AI genuinely improves user competence. To explore this, we’ll study how clinicians train with AI in simulators that mimic obstetric ultrasound -scenarios. Here, clinicians can practice - even with a potentially suboptimal AI - without risking negative consequences for patients.
This unique opportunity allows us to investigate which types of AI feedback enhance clinician expertise and which types disrupt it. By studying clinicians with varying levels of training, we’ll develop personalized feedback models that precisely support each clinician’s needs.
Thomas Bolander's project: Attention in Epistemic Planning
Attention is the mechanism that humans use to focus on specific aspects of the world. It serves as our way of managing the otherwise overwhelming amount of sensory input bombarding us and navigating the complexity of our inner mental landscape. AI systems, such as robots, also grapple with the massive influx of sensor data that needs processing, along with the intricacies of their internal representations.
Naturally, there’s interest in mimicking human attention within AI systems. Doing so could enhance robot efficiency by enabling them to focus solely on relevant information while improving their interactions with us humans, who also have limited attention spans.
This project aims to achieve just that.
Attention has already inspired various areas of artificial intelligence, including the attention mechanisms that underpin chatbots like ChatGPT. However, when it comes to AI that is both explainable and capable of planning and empathizing, we need different techniques beyond those foundational to ChatGPT. Enter symbolic artificial intelligence.
In this project, we’ll leverage symbolic AI to construct logical models of attention. We’ll develop methods to harness these models in AI systems, implement them in robots, and explore the potential for creating more efficient and social robots.
In total the Independent Research Fund Denmark (Danmarks Frie Forskningsfond) in May awarded 52 researchers 55 DKK million.
- Each year, the Independent Research Fund Denmark’s five scientific councils undertake the task of selecting the best and most ground-breaking research projects.
- The councils are composed of 75 recognised researchers with high professional expertise.
- In total, 570 researchers have applied for the DFF-Research Project2 initiative.52 applicants have been granted a total of 318 million DKK. This gives an overall success rate of 9 percent, both in terms of amount and number of applications.
- DFF-Research Project2 typically runs for up to 4.5 years and involves multiple researchers (including postdoc candidates and Ph.D. students) with a financial framework ranging from 2 to 4.3 million DKK excluding overhead.
- A DFF-Research Project2 is often characterised by being a coordinated and committed collaboration, and applicants must have significant, independent research experience at a high international level.