Focus on research: Dr Barry Sheehan, Lero
Dr Barry Sheehan is an academic attached to Lero, the Science Foundation Ireland centre for software research and a lecturer in risk and finance in the Kemmy Business School at the University of Limerick. In this interview he talks about how using AI techniques can change our attitudes towards risk.
When we ask about academics about their career paths they tend to talk about projects overseas. I think yours is a little sideways from insurance to software. How did that happen?
Indeed, I certainly consider myself a polymath – a jack-of-all-trades in the risk domain. My early academic studies were in financial mathematics (BSc, MSc), but it has always been the science of risk that has truly fascinated me. Before academia, I worked for many years in the insurance industry – in a formal actuarial function pricing motor/home insurance, and as an innovation specialist, tasked with reinventing re-insurance. Here, I learned the important role that ‘risk understanding‘ has on sustainable innovation. This led me to the Emerging Risk Group at UL and Lero to embark upon my research journey – from PhD to lecturer and beyond.
One of your recent wins was a new cyber risk tool.
Cyber risk assessment is a primary focus of mine these days. With estimates indicating that cybercrime cost the global economy just under €1 trillion in 2020, much work needs to be done in terms of cyber risk literacy. Key challenges include limited open access to cyber loss information and the rapid change at which the mode of cyberattack occurs, e.g. phishing, ransomware, Denial-of-Service attacks, etc.
In 2021 we developed and published our cyber risk classification tool, QBowTie, to support large organisations to identify, assess and mitigate cyber risks, and enable insurance companies to design appropriate insurance products. The QBowtie model can accommodate both historical data and expert judgement to score the threats, barriers, escalators and consequences for the framework.
The framework, which was tested on a city hospital in mainland Europe, produces a risk score and can pinpoint steps that can be taken to improve security measures. It also provides a practical framework that allows insurers to assess risks, visualise areas of concern and record the effectiveness of implementing control barriers.
You’ve worked on research into advanced driving assistance systems (ADAS). Are they trustworthy enough yet for widespread adoption?
ADAS has tremendous potential to reduce road traffic accidents in the coming years. Our research estimates that introducing a full deployment of ADAS across all vehicles would lower the number of road crashes by almost one quarter (23.8%). As the driving task transfers to the machine through impending autonomous vehicles,
Significant societal and regulatory challenges remain in the face of autonomous vehicles (AV). In particular, motor insurance as we know it must adapt to consider that the driving task will alternate between human and machine. Societal trust must be earned by car and original equipment manufacturers by openly communicating the privacy, ethical, and data usage policies attributed to the enabling ADAS/AV technologies.
Will data skills will become a necessity for any job in financial services and insurance?
Modern financial services (e.g. finance, insurance, fund management) require business leaders who can make the best decisions with incomplete information. These days, it is rare to see roles advertised without requirements for data management literacy. There is a particularly strong industry demand for those with both computer science and quantitative finance applied skillsets. In 2020, we launched the MSc in Machine Learning for Finance to close this skill gap and have welcomed students from diverse disciplines such as aircraft leasing, software engineering, investment banking, fund management and insurance.
Where do you see the analysis of risk going?
Artificial intelligence presents enormous potential for the risk analysis domain. From enhanced risk model accuracy to early-warning risk indicators, AI has the potential to transform the insurance business model from loss compensation to loss prediction and prevention. I closely follow the academic outputs of Martin Eling (University of St. Gallen), Mario V. Wüthrich (ETH Zurich) and Ronald Richman (Old Mutual Insure) to keep on top of the cutting edge in AI applications in insurance. Recent advancements in improving AI model explainability will be critical for the mass adoption of AI methods in the financial services industry.
Subscribers 0
Fans 0
Followers 0
Followers