AI applications and concepts
Thanks to Artificial Intelligence (AI), it is possible to find answers to social issues faster, better, and more efficiently based on data. However, what do you do with it next? Without applications, a wonderful method remains in the researchers’ drawers. Therefore, AI applications are an important part of TAISIG.
An AI application that people actually use
For example, what does such an AI application look like? Margriet Sitskoorn is a full professor at the Tilburg School of Social and Behavioral Sciences (TSB) and the cartographer of the AI application component at TAISIG. "Within the theme of Health and Wellbeing, for example, we are working with the Elisabeth TweeSteden Hospital on models for neurosurgeons and their patients with brain tumors. We already have a large number of datasets of previous patients, including their complaints, MRI images, treatments, prognosis, and clinical outcomes. Based on characteristics such as age, education, and cognitive functioning, we can classify groups and thus predict the possible outcome of treatments. To help the physician and patient make decisions regarding individual treatment, we are now designing a tool based on an even larger dataset to be collected involving multiple institutions."
It is a great example of patient centered care and personalized medicine. However, an AI application is not only at work within the themes of Health and Wellbeing, it applies to many more fields. Researcher Anouk Vermeij: "It is also valuable within themes such as poverty, logistics, crime, and education." Margriet nods. "You can apply it in large areas and at an individual level. Exactly what we will focus on depends on what the new themes of the university will be. We're discussing that."
You can apply it in large areas and at an individual level
Technology is developing rapidly, providing new AI methods, better ways of data collection, more extensive and larger datasets, and more possibilities for applications. At the same time, society is grappling with new issues such as how to deal with the growing demand for healthcare and how to encourage people to change their unhealthy lifestyles and prevent health problems. "With a Fitbit, you work on a good lifestyle yourself; by means of the COVID-19 app, you map out risky contacts between people. How do you ensure that these people actually use these kinds of tools? That is where Tilburg University can offer excellent support because of our focus on human and social sciences. ”
Confidence in AI
"Within the research for the neurosurgeon and the patient, for example, we collect images of brains," she continues. "Now, the radiologist has to look at each image personally: where is the tumor? How big is it? And how does the tumor evolve over time? It would be great if this could be automated. But then the question is whether patients trust the judgment of the AI algorithm." Vermeij: "It is necessary to involve patients and surgeons early on in the research. This can be done, for example, in ELSA labs. These are co-creation environments, developed by the Dutch AI Coalition, in which social issues are investigated. These labs cover everything from fundamental AI research to legislation, ethical aspects, and application acceptance."
It is necessary to involve patients and surgeons early on in the research
"Take the COVID-19 approach," Sitskoorn continues. "When you don't include behavioral scientists on your policy team, things don't go the way you expect. Then you can run into resistance and misunderstandings." "Plus you want to know what the effect of something is," Vermeij adds. "We can play a role in that, too." For example, in MindLabs, Tilburg University offers Virtual Reality safety training for companies such as oil refineries. In it, participants practice emergency situations without being exposed to real-life risks. Experience shows that participating companies don't want prettier or more immersive technology, but they want to know more about what the effect is. Does the approach work? Does it have a learning effect?
How do you develop an application that is demonstrably reliable, cost-effective, and quality enhancing? The answer lies in involving an entire chain in the research. Sitskoorn: "Tilburg University's Tranzo scientific center for care and wellbeing is already doing this, but for most researchers, it's a new approach to involve end users right from the start of a study. It's not a question of: can someone supply me with data for this study? It doesn't work that way. A problem is never isolated. There are always multiple groups involved, multiple types of equipment, and multiple locations and disciplines. You have to include all of that." Vermeij also sees a challenge within the university itself. "Researchers regularly find it difficult to set up a business case in the step to implementation. Then something stays on the shelf. There is still room for improvement there."
Tilburg University has a unique proposition according to Sitskoorn. "That's because of our knowledge of AI combined with our knowledge in the field of people and society." Within the university, there is still work to be done for TAISIG." Vermeij: "It would be great if we could offer researchers the opportunity to develop in the field of AI from various perspectives, for example, through training and courses." That need is visible: at least within TAISIG, a lecture series is scheduled for 2021.