Meet the new Director and Deputy Director of the Complex Human Data Research Hub

The School welcomes Professor Simon Dennis, formerly of the University of Newcastle, and Associate Professor Amy Perfors, formerly of the University of Adelaide, as the inaugural Director and Deputy Director of our new Complex Human Data Research Hub. The newsletter sat down to interview these new two key appointments to learn about them and their exciting plans.

Interview with Professor Simon Dennis

Q: Welcome to the School. It must be an interesting time for you both being at a new university and no longer being Head of School. How are you coping with all the changes?

A: Being Head of School is an intense experience. Sometimes it feels like you are being subjected to a full court press 24/7. But it is also a very fulfilling role and there was pride in what we achieved at Newcastle, but also some grief associated with leaving. Fortunately, it has been tremendously exciting getting the hub established at Melbourne. Phase one is to get to know the lay of the land and so I have been meeting with lots of interesting people. There are so many possibilities and so much energy.

Q: You’re unusual in having a very strong background in computer science as well as in psychology. How does that combination contribute to the way in which you carry out your research?

A: It’s critical in a number of ways, I think. I tend to be very focused on the mechanism of mind rather than say the characteristics of mind. My background predisposes me to think about how one could implement the computation being exhibited in human behaviour and to look for the ways in which effects in different domains (particularly memory and language) are really manifestations of the same underlying architecture. Having computer science, mathematics and physics training also provides me with a set of skills that can be usefully applied to understand cognition. My work uses probabilistic models, machine learning techniques, signal processing approaches and dynamic systems methods as ways to analyze data and to frame what the cognitive system is doing. Being fluent in these ideas also stands me in good stead in my interdisciplinary interactions. I still publish in computer science and enjoy the finality of the discipline. In computer science things stay done. By contrast, psychology is a daily exercise in teasing at one's assumptions to see where the loose threads are.

Q: Your former University of Newcastle colleague Dr Adam Osth joined us not too long ago. Are you following him? Seriously, it must be good to have him as a down-the-hall collaborator again.

A: Adam and Hyungwook Yim, who is currently my postdoc, were both students at Ohio State University as well, so it is definitely a case of getting the band back together. Adam is super smart and very skilful - and also perfectly capable of pushing back when he thinks the old man has lost it again. It’s a very creative collaboration. His 2015 Psychological Review paper, which was the culmination of his PhD work, is an absolute tour de force. His DECRA revolves around using experience sampling methods to better understand human memory - which is a passion of mine, so I'm very excited to be in the same building again.

Q: Your new role here at the University of Melbourne involves serving as inaugural Director of the new Complex Human Data research hub. That will be formally launched in due course, but what can you tell us at this point about your vision for the hub?

A: You know, there has never been a better time to be a psychological scientist. Big data approaches are changing how we shop, how we fight disease and how we date. But I don't think there is any sphere in which the potential for big data is greater than in psychology. It's actually not so much that the data we can collect now is big per se. We have been collecting neuroimaging data for a while now and it is not small. What is really exciting now is the ability to use wearable sensors, the internet of things and social networks to get an unprecedented view of the environments in which people exist. In some sense, it is the flip side. Whereas neuroimaging has given us the view inside, the new technologies are giving us the view outside - as people are engaged in their everyday activities. Rather than small slices of carefully controlled and sterilized behaviour, now we can get really comprehensive records of people's experience. That is going to change everything, both because it will allow us to build and test more comprehensive theories, but also because we will be translationally relevant in way that has not been the case in the past. Certainly, for those of us in cognitive psychology writing grant impact statements has been an exercise in dissembling and wishful thinking. That no longer need be the case. The purpose of the Hub is to ensure that the School, the Faculty and the University are not just on the cutting edge, but are the ones doing the cutting. We are planning and plotting at the moment and will be announcing new initiatives shortly.

Q: Are there particular research collaborations outside the School, whether with other parts of the University or with external partner organisations, that you are aiming to develop?

A: It is still early days in terms of mapping out the collaborative landscape and I have no doubt my beliefs about the best opportunities will update rapidly (sorry there is a nerdy Bayesian almost joke in there). Already though we have been talking with people from epidemiology and engineering and have a couple of EOIs that will be submitted soon. Outside the university there have been positive interactions with Defence Research and Development Canada, Orygen and Victoria Health that we are pursuing as well as opportunities with Carlton Connect and potentially with Northrop Grumman. We have been invited to collaborate on a CRC bid in Autonomous Systems and are, of course, planning our own CoE bid in psychological data science. I want to send a call out to Amy Perfors - who will be my Deputy Director. Amy doesn't start until early September, but has already been heavily involved in the planning and plotting.

Q: You are known as a memory researcher, among other things, but you also have a deep interest in what is variously known as ‘ambulatory assessment’ or ‘ecological momentary assessment’. Can you tell us a little about that and what you think it can contribute to psychological science?

A: Also known as experience sampling, the quantified self and lifelogging. Would be nice to have A term for this kind of work, but no doubt any attempt to impose a standard term would be as successful as such endeavours typically are. I guess I have already answered this question above, but perhaps if I can expand with an example from my own work to give a sense of why I am so bullish. One of the fundamental tenets of memory is that performance depends on experience. This statement is true in multiple senses. Firstly, in order to retrieve memories we need to encode them. But we LEARN to encode and experience determines how we go about it. Harry Barrick was one of the primary figures in memory research in the 20th century (he lived through a great portion of it). I asked him a few years ago what he thought the main result of the century had been. He replied that it was the understanding that there is no such thing as a nonsense stimulus. When Ebbinghaus first started experimenting with human memory he used randomly constructed syllables (like YBW or KWJ) in an effort to control for people’s prior experience. In the end though it was never entirely successful as people are quite ingenious at constructing meanings. Perhaps YBW means “why be W when you could have been just you?” How people encode depends on their prior exposure to stimuli and to encoding strategies.

Secondly, you can never isolate people from interference. After encoding efficiency, the next most important determinant of memory performance is interference – both from the experiences that occur between encoding and retrieval (retroactive interference) and from the experiences that occur before encoding (proactive interference).

All of this creates a problem for memory research though. Typically, we bring people into the laboratory, show them some stimuli and then ask them questions about those stimuli. We know that their prior experience is critical determinant of performance, but we have only very crude measures of that experience. We might record the frequency that a particular word appears in language – assuming that it would have been about the same for that individual – or we might ask them to write diaries or rely on the stories of family members and friends – all products of the memory system themselves.

Experience sampling promises to change all that. For the first time, we can sense what an individual is experiencing as they experience it. We can create a veridical and ubiquitous record and use that to inform our models of memory processes. In our experiments, we have people collect images, audio, GPS and accelerometry data for two weeks and then bring them into the laboratory a week later. We present images and ask them to indicate when the images were taken. Using hierarchical Bayesian methods, we can not only establish how people go about the task, but also make estimates of the extent to which they encode visual, auditory, spatial and movement information - in their everyday lives. In the immortal words of Dr. Phil, we are getting real.

Q: Do you have a paper that illustrates some of your recent work? What is its main contribution and what questions does it raise for future research?

A: In the manifesto above, I may have given the impression that somehow experience sampling work is in competition with laboratory or neuroimaging approaches. However, I want to emphasize that this is not the case. The main methodologies in psychology have complementary strengths and weaknesses and invariably the best science is done by drawing liberally from all of our tools. To emphasize the point, I have chosen a paper that was published in PNAS in 2015 in which we used a combination of experience sampling and fMRI to ask questions about how the recollective experience of time and space is encoded in the brain. I hope you enjoy it and please feel free to contact me if you want to discuss memory, the paper or anything to do with the Complex Human Data Hub.


Interview with Associate Professor Amy Perfors

Q: You’ve just joined the School of Psychological Sciences from the University of Adelaide. What are your first impressions?

I'm pretty excited! At the time of this interview I've technically not officially started yet and have only been in a few times, but each time has confirmed that I made a good choice in coming here. I've appreciated the breadth of people doing interesting work, the enthusiasm for ideas, and the fact that there is a very rich intellectual community around Melbourne. Also a big fan of the view from the 12th floor.

Q: You have worked in psychology departments since you arrived in Australia but your background is unusually broad. I gather your undergraduate and master’s degrees at Stanford were in symbolic systems and linguistics rather than psychology, and your MIT PhD was in brain and cognitive sciences. How does this breadth contribute to your research and teaching?

I think it reflects the fact that I have a strong tendency to find almost anything interesting. On the plus side, it means one of my strengths is an ability to draw connections between lots of different areas.  On the minus side, it means it's easy for my attentions to get too dispersed if I'm not careful. That said, although it's pretty diverse, all aspects of my background do reflect my core interests: how ideas and concepts are represented, how we use them to make decisions in the world, and how communication and our social world shapes and is shaped by them. Pretty much everything I do is related to those things in some way, and I think that's the principal component that links everything that I do.

Q: Your research falls within the broad domain of cognitive science and mathematical psychology but it also has a strong developmental dimension. You have studied language acquisition and children’s reasoning processes, among other things, and learning appears to be a major theme. How important to your work is this developmental or dynamic perspective?

The developmental perspective is important in two ways, I think. First, it's where I started from -- my earliest work was in a developmental lab focused on language acquisition -- which means that it has served as my intellectual anchor, so to speak. The practical effect is that I have a tendency to always wonder about any new idea in terms of what it means about our early input or predispositions. Second, it's given me a real appreciation of how you can't really completely separate developmental work from work on adults. Even though nowadays most (though not all) of my research is on adults, I think that it's vital even for theorists about adults to understand people's cognition through the lifespan. What we know about adults has to constrain developmental theories because adults are the endpoints of those; similarly, any theorist about adults should be thinking about what developmental process might lead people to that adult state.

Q: Your new role here at the University of Melbourne includes being inaugural Deputy Director of the new Complex Human Data research hub. Can you reveal any of the early plans you have for the hub and your own work in it?

Honestly, my first and main priority with the hub will be to talk to lots of people, both within the school and throughout the university, to get a much stronger sense of the relationships and skills we have to build on in creating it. I think the hub will be successful to the extent that it reflects the actual interests and enthusiasm of folks on the ground rather than being imposed top-down by myself or Simon. That said, my first sense is that there's a lot of space to do this. I'm pretty excited about the idea of using mathematical and computational techniques as the methodological "glue" holding together several styles of approach and methods of analysis, from experience sampling to computational social science to more traditional lab-based cognitive psychology. So I'm hopeful that we'll be able to construct something reasonably coherent around those axes. If that's the shape it ends up taking, I think my own work will overlap non-trivially with all of these components in some way.

Q: Are there particular research collaborations inside or outside the School that you intend to cultivate?

For sure! That said I know that whoever I name here, I'll inevitably think of other people I should have named -- and I'm also confident that I will find people to collaborate with that right now I'm not aware of at all. So, since discretion is the better part of valour, I'll avoid naming any names here and simply say that I'm excited to collaborate with lots of people, from those already studying cognition and decision-making to people studying social networks, and potentially extending outside of the school (e.g., mathematics, computer science, linguistics, behavioural economics).

Q: There is a rumour floating around that you are an avid rugby fan and (former?) player. Is there any truth to this rumour?

Ha ha - the rumour is correct! More of a player than a fan, actually. I played for about 15 years, starting at university, until I had to stop due to injury several years ago. Since then I have found it pretty jealousy-inducing to watch rugby games because I wanted to be out there myself! I'm still searching for something that fills the same niche. Turns out that I really liked violently tackling people into the ground...

Q: Can you recommend a recent paper that demonstrates the sort of work you do? What makes it important, and what are its implications for research in your field?

One of the ones I'm most excited about is still under review [ can be found here]. In it we use mathematical and computational analyses, and a few experiments with adults, to explore how information changes as it is passed through a population of people. We show that, even if you assume that everyone involved is "rational" (i.e., Bayesian) then as long as they have different priors, the information that survives is disproportionately shaped by the people who have very strong priors. Colloquially, what this means is that extremists -- people who are more unwilling to change their views in response to data -- shape the overall conversational dynamic more than they should. The effect can be quite large: even if there are a small minority of these people (like 5%) they can dramatically shape what gets talked about and passed around, way more than you'd predict based on their proportion. This has implications for things like language evolution, but also for the dynamics of communication in the world.