In Watson, my team investigated what a potential personalized television future could look like. This was quite a challenge, considering we didn’t have an online video platform that we could use for tests. Also people weren’t ‘ready’ for such a personalized service. To get an idea of how such a service should function, we conducted several experiments and studies to find out why people watch television. This way we could translate that core feeling of watching TV into a futuristic concept.
Among the experiments, I would like to highlight one of them: the human algorithm. At a certain point in the project we wanted to test out some insights and create a machine that could predict a ‘perfect TV watching evening’ for our test users. However, as stated above, we didn’t have an online video platform or whatsoever to create a good-looking player. Neither did we have the time or the resources to invest in a large-scale recommendation engine. Mainly because at this stage, we just wanted to know whether we were going in the right direction. Our solution was pretty simple. Instead of a machine, we asked Jo Martens (a specialist in TV-programming), who works for the VRT research center to create a personalized watching scheme for our test users. This was based on their agenda, some demographic data and profiles that we had given them. Each evening we would send out the personalized scheme to our test users and call them up late at night to talk about the scheme. We asked whether they had followed it and why they did or did not. We noted this information in a diary we kept for each user. Jo could use that information for the next day. Even though this seems like a small experiment, our research gained a huge boost in finding out why people watch television.
Other experiments included emotion tracking and recognition, physical activity monitoring and context analysis of our test audience.
In the end, we created a prototype that would incorporate all of the findings and thus created a, what we believe to be, ‘perfect’ TV evening. However this is just a theoretical product: we didn’t create a recommendation engine just yet, but we are well on our way.
I was involved in most of the experiments as a researcher, designer and developer to set up the tests. This was a two-year project of which I was project lead during the second year. Furthermore I have designed the interaction flow and user interface of the final prototype.