Connecting European Neuroscience

The Brain Conference on Computational Neuroscience of Prediction: Q&A with the co-chairs

The application procedure is now open for the next Brain Conference, which is taking place 15-18 April 2018, in Rungstedgaard, North of Copenhagen, Denmark. Co-chairs Peter Dayan and Kerstin Preuschoff talk to FENS about the topic and what the attendees can expect.

*The spring Brain Conference traditionally follows the theme of the Brain prize winners of the previous year. In what sense does the conference Computational Neuroscience of Prediction fit into the theme of reward?

First, although the topic of the Brain Prize was indeed reward, much of the work that was recognized in the award actually concerned the provenance, nature and effects of predictions of future reward. In fact, the more deeply you look at reward, the more intertwined with prediction it seems - organisms that don’t predict well typically don’t end up with much reward. Prediction is a substantial and fascinating topic – studied richly across modern cognitive, behavioural  and computational neuroscience – the conference will touch on many elements of this. Second, gain cannot properly be defined in the absence of pain; so it would be invidious not also to study punishment.

*Brain Conferences are also traditionally multidisciplinary. How are you bringing multidisciplinarity to this meeting – don’t theoreticians and experimentalists work together routinely in basic research areas like prediction?

Yes: one of the very best things about prediction is its inherent multidisciplinarity. Pick a discipline – economics, statistics, control theory, AI, psychology, neuroscience – and one way or another it will have studied prediction and developed its own approach and methods to describe it. In the Computational Neuroscience of Prediction, we will draw on all of these fields, and attempt seamlessly to integrate their ideas. The speakers who are attending are paradigmatic multidisciplinarians, each presenting a unique approach towards this integration – so we are optimistic for healthy interchange and debates.

*Some say that the brain is simply one big prediction machine. How solid is this concept?

We knew you were going to ask that question! But seriously, prediction is central to the smooth functioning and success of any system – particularly one doomed by slow and variable sensors to live noisily in the past, but to seek to control a dynamic environment. From a theoretical point of view it’s the best strategy in a non-random environment, and, as our colleague, Read Montague, has compellingly pointed out, it would be surprising if millions of years of evolution hadn’t worked this out!  However, perhaps surprisingly, the centrality of prediction implies a danger of vacuity – the scientific challenges then become asking which of many possible things are predicted; how are those predictions made when optimal prediction is wildly computationally intractable; how do we tune our predictions in the light of the actions that they mandate; how do we balance out the task of building better predictors with that of choosing better actions. All of these will be aired at the conference.

*What key breakthroughs are needed to help understand the brain mechanisms of prediction?

There are many challenges for understanding prediction, and each speaker will have his or her own ordering. Some that we think are particularly pressing include working out what to predict in the first place, balancing or integrating different methods of prediction (model-based and model-free to adopt a prevailing distinction), determining the many loci in the brain that predictions are made and understanding how they are represented.

*What are some of the less obvious issues around prediction that are likely to be discussed at the conference?

We chose speakers based on what we predicted they would talk about. So it’s hard to turn around now to expect the unexpected… However, we’re sure that we can predict that there will be unpredicted things – with the wonderful and broad panoply of investigators at the tops of their games. We chose speakers from a wide range of disciplines who like to think outside the box and identify new areas of study. Based on past experience, the ‘less obvious issues’ tend to be appear in highly multidisciplinary environments such as the one we are trying to create with this conference. If you really pushed us: the link between prediction and attention seems to be a common theme – it is definitely something to look out for.

*What do you think young scientists at the beginning of their careers might gain in particular from the meeting?

It’s a rare opportunity for them (and us) to interact with a rather unique mix of scientists. To the chagrin of the organisers as we sweat over the programme, the most beneficial outputs will almost certainly be the informal connections and information flow that the meeting will provide. The wonderful setting is conducive to free-flowing discussions; participating in, and even just watching, this is invaluable.

More on The Brain Conference here.

Start your application here.