Where is Neuroeconomics going? (2/2)

Elliot Ludvig is an associate professor in the psychology department of the University of Warwick, who coordinates the MSc Behavioural and Economic Science, and teaches the most amazing course there: Neuroeconomics. In this part of the interview, he’ll give his view on neuroeconomics as a field, its limitations and what he thinks its future might look like. Disclaimer: everything mentioned in this interview is, of course, an opinion.

Q: You mentioned cool papers currently coming out with regards to model-based learning and learning from experience. Are those topics the way forward for neuroeconomics? Is that how you see the future for neuroeconomics? That is one of the current lines of research. Like I said before, there is the move towards computational psychiatry, which is really interesting, but mainly useful if you want to look at disorders. That doesn’t help us study “normal” people. It doesn’t teach us how to make most people make better decisions, which is what most of behavioural science does. Such as promoting saving and recycling. Is neuroeconomics teaching us how to do this better? Or is behavioural science better at teaching us how to do this better? Neuroeconomics will hopefully help us build better theoretical frameworks for how people actually make decisions.

One thing that neuroscience takes much more seriously, which economics and psychology don’t, is that we are animals. When you are dealing with the brain substrates, you become very aware that you are just a piece of meat that moves around in the world and engages with things. We are just a collection of biological tissue. We are not this idealized, rational creature. We are built out of this messy, wet combination of neurons that we have gotten from our ancestors just like all the other creatures around us. That must be very humbling. It is very humbling. It is a reminder that animals are a lot cleverer than we give them credit for, and we might not be nearly as clever as we give ourselves credit for. It also reminds us of how most things we come up with are complete confabulations. We know this from brain lesions studies. We come up with stories as to why we did something, but this is all completely made up. We have been re-emphasizing how irrational we are. Build-up of non-optimal things that have survived and developed within constraints. I think that this is very insightful. I think this adds a significant layer to behavioural science. This is what I think is important about the brain-based perspective, in addition to the theories it inspires.

Q: Where do you think the field will be in 1 year, 5 years or even 10 years from now? Take an educated guess, I’m just very curious on your perspective. This is a good question, but difficult to answer. I am aware of the field, but I am also aware of the strong limits on our ability to predict the future. As we know from other areas in forecasting (laughs).

On one the hand I can say that we are 5 years away from finding the ultimate neuroscience theory. If we only take reinforcement learning far enough, into AI… But I think that is too much.

Q: What has been coming up trend-wise, in research? And where do you think that is going? If it’s going anywhere at all. Neuroeconomics is constrained by its tools and the tools have been getting a lot better. We have recently developed a version of ultrasound which can actually stimulate deep-brain, rather than TMS which could only stimulate right on the surface. This will enable us to engage more in causation.

What I do see happening is attempts at making it more commercially viable. What does exist in neuroeconomics, people will use to predict and influence consumer behaviour. More into the behavioural domain. That is one direction, I already mentioned computational psychiatry. I also see the big open questions, there are still big fundamental questions left unanswered. What are the core computational parts in the brain? How does the brain compute prediction errors? We have this idea that t-learning works well for a lot of situations, but there are limitations. The consilience with deep-learning will happen as well. Relating AI to the brain. I think we will continue to grow into bigger and harder questions. Questions regarding social neuroscience and moral decision-making. I think those will be the directions in which people will ask some tougher questions. In some ways, this chases the psychology. It looks at what is happening in other areas and integrates them. But it also drives things, through the usage of tools.

I think the coolest stuff still tends to come out in the intersection where people focus on these reinforcement models. It's just fascinating to see how people learn, how they use past experiences, how they build models of the world and how they make inferences. That is where the coolest papers are happening now.

Those are my vague set of predictions, which I can look back on in 5 years and I say I was right. Or just not too wrong. Exactly, they aren’t nearly specific enough to have pre-registered specific metrics that tell me I’m wrong. I hope that some of the advances in other nearby areas will strongly, more strongly infiltrate neuroeconomics. Because I think that will clean up and improve the quality of the science to a large degree. Because that is something I worry about. How much would this stand up to replication? To robustness or to other analyses? Sometimes yes, it will, but I do worry a little bit. So that is a methodological point in which I hope the field continues to evolve, to yield better answers.

Q: Given that my readers can hardly be expected to already be full-fledged neuroeconomists, what would you recommend for them to get into the field? Are there any must-reads for example? A really good introduction is chapter 7, in the New Mind Readers, by Russell Poldrack. That chapter is all about decision neuroscience. The entire book is a good introduction as it remains very skeptical. It shows what we know, what we can already do, but also focusses on limitations. And of course, my own paper on reinforcement learning with Bellemare from 2011. It shows all three sides: empirical, computational and medical. And of course, they should read your blog. Laughs. Shameless promotion. And they should come here and take my class. But all of this is academic. They aren’t exactly popular science-type introductions.

Q: Do you think neuroscience lends itself to pop. science introduction? Well Paul Glimcher had this book Decisions, uncertainty, and the brain, which wasn’t bad as an introduction, but it’s dated. It’s from 15 years ago and the field has progressed massively since. Dave Redish also had his book, the mind within the brain, that had a lot of neuroeconomics in it. So there have been a couple of attempts. I don’t think there is a real Predictably Irrational for neuroeconomics, a book that gives you a warm welcome and tells you everything you need to know if you know nothing. But maybe there is and I just don’t know it. Maybe I should write something. I mean the field is open. You heard it here first: Elliot is coming out with his own book! I think it might be hard to write something really accessible, in such a way that everyone wants to learn more about it. Something like Thinking Fast and Slow, needs to be written for neuroeconomics. It's a bit harder to read, but you can get it, it’s not that hard. I’m sure that if you give me more time, I can come up with something that does a great job of introducing neuroeconomics. But those are not bad!

This is the end of the last part of the interview. Thank you very much Elliot for taking the time to discuss neuroeconomics. We can look back in 5 years and see whether your predictions were (somewhat) correct! References made throughout: Crockett, M (2012) Beware Neuro-bunk. TED Talks. https://www.youtube.com/watch?v=b64qvG2Jgro

Glimcher, P. W. (2004). Decisions, uncertainty, and the brain: The science of neuroeconomics. MIT press.

Ludvig, E. A., Bellemare, M. G., & Pearson, K. G. (2011). A primer on reinforcement learning in the brain: Psychological, computational, and neural perspectives. In E. Alonso, E. Mondragon (Eds.), Computational Neuroscience for Advancing Artificial Intelligence: Models, Methods and Applications (pp. 111-144). Hershey, PA: IGI Global.

Poldrack, R (2018) The New Mind Readers: What Neuroimaging Can and Cannot Reveal about Our Thoughts. ISBN: 9780691178615

Redish, A. D. (2013). The mind within the brain: How we make decisions and how those decisions go wrong. Oxford University Press.

Schultz, W., Dayan, P., & Montague, P. R. (1997). A neural substrate of prediction and reward. Science, 275(5306), 1593-1599.

Sutton, R. S., & Barto, A. G. (2011). Reinforcement learning: An introduction.

Sutton, R. S., & Barto, A. G. (2018). Reinforcement learning: An introduction. MIT press.