Last week I had the chance to talk to Eva Blomqvist about the role of Semantic Technologies in privacy-aware environments.
Thomas: You are Associate Professor at Linköping University with a strong legacy in Semantic Web technologies, ontology design and a particular interest in decision support systems for the security and e-health domains. How do semantic technologies contribute to this area? What are the challenges from your perspective?
Eva: Data management and data integration are huge challenges in both the security and e-health domains, there are actually similar challenges in both of them. For instance, police analysts today have to sift through large amounts of data in order to find the small pieces relevant for an investigation, and even only considering internal data sources a police department can have hundreds of legacy databases, from which data needs to be collected and combined. In e-health it is often about making sense of sensor data, measurements, and patient self-reporting, together with health record data. Semantic technologies can help to make sense of all this data and to integrate the pieces in a meaningful way. Considering the broader field of AI, we have come a long way using machine learning in the past decade, however, I believe that the really big challenges can only be solved by combining machine learning with ontologies, or other semantic technologies, in order to provide hybrid reasoning strategies and more transparent results. Transparency is a big issue for both these domains, in particular.
Another common challenge is the analysis of near real-time data, such as streams of vehicle movements elicited from traffic cameras that are used in certain police investigations, or patient monitoring by the use of sensors and IoT devices in e-health. This data can usually not be stored in its entirety, but reasoning needs to happen "online" while receiving the data, and only the important pieces, related to the current investigation or that is relevant to the current status of the patient, should be kept and further dealt with by the system.
Thomas: Support systems for security and intelligence analysis very often unfold a tendency towards surveillance and centralized control of data. How can semantic technologies contribute to data sovereignty and protect us from excessive data exploitation?
Eva: On one hand, both in security and e-health there are already strong privacy and access control requirements, both in terms of legislation as well as user requirements. In this case, semantic technologies can help to be able to set access restrictions at the right level of granularity, based on the actual data content, and to understand what pieces of information need to be treated carefully to protect privacy. Overall, our projects have always been done in collaboration with legal and ethics experts, in order to make sure that the technologies are used in the best possible way, i.e. to serve all citizens without excessive surveillance. I think this is a very important aspect! One concrete example is from our latest e-health project where one of the challenges was to asses the patient's situation completely without video or audio surveillance, precisely for preserving their privacy in their own home.
In general, on the other hand, I think we also need to move more towards stronger ownership of your own data. Today, we give up too much information online, without having any control over where this data ends up. There are interesting initiatives to change this situation and allow us to take back control over our data, such as the SOLID project ran by Sir Tim Berners-Lee, which builds on semantic technologies. It would be very interesting to in the future look more at how such initiatives can work together with, and hopefully, actually empower law enforcement and other societal functions, rather than hinder them.
Thomas: Thinking ahead of the obvious next developments, is there a topic or development which you want to see discussed, even if it’s science-fiction?
Eva: I think the most interesting general research direction at the moment is the integration of machine learning and knowledge representation or for instance the combination of deep learning and knowledge graphs to be more concrete. It is not yet clear exactly what will be the silver bullet that generates the next big leap in AI, but I am absolutely sure that it will happen through such integration. This has of course been discussed at the conference already in the past few years, but I would love to see some really novel ideas and innovative work in this direction at this year's conference, and especially adding some near real-time data streams and stream reasoning into the mix.
Eva Blomqvist is an Assistant Professor, in the MDA lab (HCS division) at IDA, Linköping University. Research is focused on decision support systems, in particular for the security and crisis management domain. Particular interests include Semantic Web technologies, ontologies and ontology design patterns etc.