Dhaval's passion for Responsible AI shines through

My research focuses on explainability, trustworthiness and equitability in AI and the responsible use of Deep Learning. For example, because of the number of parameters used in a system like ChatGPT, it gives the user the impression that the system understands what you are saying. So, when you ask a question, it leads you to believe that it understands the context of the question and gives you an answer which seems to be sensible, logical and often helpful in a number of scenarios.

Although AI can appear to be ground-breaking, it can be misused or behave unpredictably. Hence, my research delves deeply into constructing AI systems with responsibility at the forefront. It's pivotal to establish robust guardrails, ensuring that these platforms not only function effectively but also act responsibly, elucidate their decisions and recommendations, and consistently earn users' trust.

My research is about being responsible for the power AI brings and addressing the challenges. Working with social scientists, we ask the question, how do you make AI systems that are equitable? How do you bring in social aspects, ethics, and the law when building AI? And, how do we deal with bias which might be inherent, when you're building AI systems?'

We work on a number of health issues with a focus on self-management and support for patients and caregivers in managing chronic conditions such as Asthma, and Diabetes. We use the Internet of Things sensors, self-reporting and AI-based monitoring, recommendations and predictive analytics.

For example, I have also worked with health scientists in the care of Asthma patients, promoting self-care. We look at how AI can use the data from Indoor Air Quality Sensor Devices to offer guidance on the kind of activities patients can safely do and the kind of activities they might want to avoid.  By using data on air quality, patients can avoid situations that could exacerbate their condition.

We also collaborate with city councils across the UK and Europe to develop AI-based frameworks that assess interventions aimed at enhancing air quality in their cities. For instance, we have partnered with Bradford City Council to gauge the efficacy of their Clean Air Zone. This involves augmenting the monitoring of air pollution through AI and IoT, complemented by spatial-temporal AI research to bolster such monitoring.

Explainability and transparency are crucial in ensuring Responsible AI.  They make systems more accountable and interpretable.  In the detection of tumours, for example, can AI systems explain to a doctor why certain images are more likely than others to show breast cancer? Can the AI system explain its decisions? Can AI explain the rules it has developed internally? What does it learn from the data provided? If it can explain, we can 'train' it more accurately and think about how to include different sorts of data.

My research group specialises in a distinct strand of AI known as Neuro-Symbolic AI (NS-AI) to achieve 'explainability' in AI systems. NS-AI integrates neural network approaches with symbolic reasoning to enable AI models to provide more interpretable and logical explanations for their decisions.

There's a lot of work going on to ensure that AI is trustworthy and safe. At the University of Hull, our research delves into the vulnerabilities of machine learning algorithms, exploring scenarios where they might be misled into rendering erroneous decisions. If they can be fooled, it undermines the safety of AI. We are working on techniques to estimate the safety of machine learning classifiers.

I get excited by societal challenges, especially in the health and environmental fields. My drive stems from a desire to harness the power of AI to tackle and find solutions to these challenges.

I wanted to come to the University of Hull particularly because of the potential for working with the Hull York Medical School to expand the use of AI to address health challenges. Additionally, the University boasts an impressive research track record in environmental fields and Net Zero, which aligns with another focal area of my research.