UCL Research Domains


Together We Create - Episode 3

Towards meaningful interdisciplinary working in digital health research

In this episode, we are joined by Dr Olga Perski a collaborative social researcher and practitioner health psychologist working in digital health.


Lili Golmohammadi  00:04

Hello and welcome to Together We Create, a podcast about collaborative social research. My name is Lili Golmohammadi, I'm a collaborative researcher working across design technology and social research and the final year PhD student at UCL. In each episode, I'll be talking to an early career researcher at UCL to find out more about how and why social researchers collaborate with engineers, scientists, health practitioners and designers and hearing about their research stories and top tips as we discuss the benefits and challenges of taking a multidisciplinary approach. 


Lili Golmohammadi  00:41

In this episode, I'm joined by Dr. Olga Perski, a collaborative social researcher and practitioner health psychologist working in digital health. Olga was training to be a classical musician, but in 2014, following an internship in a Stockholm clinic, she joined UCL to study for an MSc in health psychology, and then a PhD on understanding engagement with health and wellbeing smartphone apps, and in 2018, Olga took up a research post in the UCL Tobacco and Alcohol Research Group in the Department of Behavioural Science and Health. So welcome, Olga. Thank you for joining us today.


Olga Perski  01:15

Thanks so much for having me.


Lili Golmohammadi  01:17

So your work brings together behavioral scientists and technologists, and it's a relatively new field. Can you tell us a little about what led you here?


Olga Perski  01:26

So I'm Swedish. And before I moved to the UK to study, I was very interested in how to support people with addictive problems. So particularly related to substances such as tobacco and alcohol. And being Swedish, I saw that quite kind of early on, or Sweden seem to be quite an early adopter of using computers and the Internet to deliver interventions, particularly for or in the mental health area. So Internet Based Cognitive Behavioural Therapy. So I kind of at the same time, as I started an undergrad in psychology and philosophy in the UK, I then kind of tried to pursue this interest and did an internship in the summer back in Stockholm, working at one of the first E-Psychiatry units, where they had, I think it was about four different randomised control trials going on at the time, evaluating the delivery of Internet Based Cognitive Behavioural Therapy. And for my Master's, when I was looking at kind of programs, I was really interested to see if I could combine kind of my interests in addictive behaviours and technology. And then I found that at UCL, Professor Robert West was heading up a few projects that involved smartphone apps to deliver Stop Smoking Interventions. So that really caught my attention and kind of made me apply to the course. And then I was lucky enough to be able to do my project with him and actually, then also, my PhD.


Lili Golmohammadi  02:54

That's quite an interesting trajectory. I had no idea that Sweden were early adopters in this area, do you know why that is?


Olga Perski  03:02

Well, I suppose it is a country that's had quite good kind of technology infrastructure, and possibly also given that it's a smaller country was able to kind of adopt some of these technologies within healthcare a bit earlier than some other countries. But yes, I would say that kind of Sweden and well the Scandinavian countries and also the Netherlands were quite early adopters of using technology to deliver psychological interventions.


Lili Golmohammadi  03:26

Were you ever skeptical or concerned about the potential of digital technology to be of use?


Olga Perski  03:32

Absolutely. So actually, that was kind of one of the key reasons that I was interested in using technology to deliver these types of interventions, because we do know from a wealth of research that it's sort of in the interaction with a therapist or a healthcare professional, what we sometimes call the therapeutic alliance, that is very important for for outcomes and kind of improving health. So that was sort of one of the areas that led me to want to pursue a whole PhD actually looking at kind of how people interact with these technologies, and then how that relates to outcomes, and also trying to think about identifying different design elements that can influence how people interact with these technologies. But yes, I would say that I approached the area with some healthy scepticism.


Lili Golmohammadi  04:21

And do you feel like that's still healthy, but less less, so less levels of scepticism at the moment?


Olga Perski  04:28

Well, I think it's an important part of research to maintain kind of that critical thinking throughout and in my particular area it is very challenging, particularly when we're thinking about trying to sort of work out the effectiveness of something compared with something else. And whether or not it's worthwhile for the healthcare system to invest in these types of technologies, because what tends to happen is that people who engage more with these types of interventions have better outcomes on average, but given that people aren't necessarily kind of randomised to their level of interaction, it means that it becomes very difficult to think about causal inference, essentially. So engaging with this particular technology actually directly leads to the outcomes, or was it something about the person themselves. So it's quite a challenging kind of research area as well from that perspective.


Lili Golmohammadi  05:20

So that brings us as base to your current work, which is a collaboration with computer scientists researching developing smartphone apps and chatbots, to help people stop smoking and reduce their alcohol consumption. So what's the focus of this project and what does it involve?


Olga Perski  05:36

Yeah, so I'm currently working across a few different projects, all involving different forms of technology, such as smartphone apps or virtual reality, to try to deliver interventions mainly for smoking cessation. One key project that I'm working on at the moment that I'm very excited about is in collaboration with two scientists from UCL. So they're kind of experts in data science and engineering and we're trying to see whether we can use data from smartphone apps to try to predict when somebody is about to have a smoking lapse. And the next step from that would be to then try to see if we can anticipate these moments and deliver an intervention message at the right time for the person.


Lili Golmohammadi  06:22

But you've also been exploring chatbots, from what I understand, can you explain a little bit about chatbots, and what they are and how they work and how they work in this context?


Olga Perski  06:32

Yeah, so I suppose chatbots are either kind of text based or they can also involve kind of other modalities, but they're computer programs that can have a two way interaction with the user. And in the particular context that I've worked with them is that so in collaboration with the originator of the Smokefree app, which is one of the world's most popular apps, so it's got about 4 million downloads to date. And the developer was very keen to implement the chat bot within the whole ecosystem of other functionalities. And we then kind of did a study to try to understand the effect of kind of adding this chatbot to the other components in the app on things like user engagement, and then also smoking cessation. And just to tell you a bit more about the study that we ran. So what's quite exciting about this area as well and particularly in collaboration with the Smokefree app is that because it's got such a large user base, we were able to run an experiment and recruit over 57,000 users in just about three months. And we then randomised users to either kind of receive the app as it was, or with the addition of this new function, which was Quit Coach, as it was called. And we found I mean, what was quite exciting as well, is that we saw quite a huge impact on user engagement. So we saw that it led to just over double the rates of engagement as kind of with the standard version of the app. So that was quite exciting preliminary findings.


Lili Golmohammadi  08:11

Yeah, that is exciting, yeah. And then I suppose the next stage is to see whether that impacts on people giving up and is that the idea?


Olga Perski  08:22

So yes, we also did look at some short term quit outcomes. So we looked at what happened after a month and we also saw that it had a positive impact on quit success. But also kind of in the context of this type of research, because often, when you would run kind of a clinical trial, you probably would want to try to get some biochemical verification of smoking status at the end, so we didn't do that here, for example - it was all self-reported. So even though kind of this type of real-world research is very exciting, and we can reach a lot of people, there are also some constraints that may then impact on kind of the, how we interpret the findings from those trials. So, we actually did a qualitative study as well, and spoke to a few users to try to understand kind of what happened over a couple of weeks of use. And what was quite interesting is that people obviously kind of well knew that they were interacting with a bot, but the people still felt that they were kind of creating some sort of bond with the bot, and that the bot was kind of there to support them. But and then just to highlight kind of one other interesting thing as well, is that because I suppose in the interaction with a human, you do, hopefully, expect the relationship to be two way in that, you know, if you've got a friend, you've got to ask them how they're doing and it's not just about how you're doing, but our users kind of seem to well, they they kind of highlight that they actually thought it was quite nice that the interaction was one way [laughs] so it was quite different to interacting with kind of a friend or a family member who might not want to listen to kind of them craving a cigarette and not knowing what to do about it. So, it does seem that even though it's kind of well, supposedly creating this two way interaction it might be but it's not not necessarily very sort of reciprocal.


Theme music  10:11


Lili Golmohammadi  10:18

So, in this podcast series, we're exploring different ways of doing collaborative social research. And so, I want to talk to you now about how your research actually works. So how do you bring the ideas and methods of health psychology, behaviour change, and computer science together in your work?


Olga Perski  10:37

I was fortunate enough that during my PhD, which was looking at kind of understanding how people engage with apps for health and well being, I was then working with supervisors across these different disciplines. So I had supervisors specialising in behavioural science, but then also Professor Ann Blanford, who's at UCL and focuses on human computer interaction, so quite early on in my training, I realised that it was important to draw on literature and kind of methods from these different fields, saying that, it can be quite challenging as well to bring in these diverse literatures and try to make sense of the literature or but also, for one's own research, I think in terms of both in terms of kind of scholarly identity, but then also like, where should I... or what's the natural kind of audience for my work or journals to publish in or conferences to go to. So I don't think that I necessarily have sort of plants. But what I've tried to do, or I think what I did try to do in my doctoral work was at least to read quite kind of widely and bring in different methods but then I had to be, I think, quite strategic about sort of where I was kind of directing my work and how I position myself. So I think I did choose to go a little bit more kind of towards behavioural science and health audiences rather than human computer interaction.


Lili Golmohammadi  12:13

And what drives that decision in the end, do you think?


Olga Perski  12:18

Well, it's interesting, because some of it was actually through trial and error, kind of submitting my work to, for example, human computer interaction conferences without having much success, then kind of made me realise that it probably wasn't written in a way that appealed to that audience or kind of the questions that I was asking, even though they were perhaps very novel to kind of a health audience or behavioural science audience, they weren't necessarily seen as novel to this other audience. So I think that that's one aspect. The other aspect, I think, is kind of coming back to sort of the research questions and the purpose. So I suppose in the health fields and behavioural science, we're often asking questions around kind of the effectiveness of interventions, which then requires the use of experimental methods, for example, in order to answer those questions, whereas kind of in other disciplines, they might - the questions might be very different. 


Lili Golmohammadi  13:18

Could you give some examples of like a couple of methods from health psychology, and from behaviour change and computer science and how they mesh together?


Olga Perski  13:28

I think working with technology and kind of drawing in some methods from human computer interaction, some of the qualitative methods that are used a lot would be around understanding people's interactions with technology, but using qualitative methods. So it could be diary studies, it could be the Think Aloud technique, for example. So I think bringing those methods in to health psychology has been very useful. But then also, I think what's important to mention as well is that it's kind of in this new field of digital health and thinking about how can we evaluate these technologies, given that they're kind of rapidly evolving and kind of using iterative design, there has been a lot of novel methods that have been developed in the last decade or two decades, more novel frameworks that are drawing on kind of behavioral science and engineering. The Multiphase Optimisation Strategy Framework, which kind of is suitable to kind of interventions with different components that can be delivered in a myriad of ways. So before kind of putting everything together into a package that's evaluated in a clinical trial, sort of highlighting the importance of trying to understand how those different components work sort of separately and synergistically, and then try to kind of optimise the intervention before it's evaluated in a randomised control trial. And then kind of other kind of evaluation methods that have come out of this kind of interdisciplinary work between health Psychologists and engineers and statisticians is also the micro-randomised trial, for example, where instead of having two options that control an intervention in the micro-randomised trial, people would actually be randomised over and over two different intervention options over the course of the trial. So more about understanding what works in a given moment or given context, rather than sort of delivering something at baseline and then evaluating it later on.


Lili Golmohammadi  15:31

Okay, so there's lots of methodological opportunities coming through as technologies develop, and all these fields develop alongside that. So is there any feedback loop as well from this kind of application to the development of behaviour change Theories as well?


Olga Perski  15:50

Yeah, so I think that's, for me, at least one of the most exciting things about technology or kind of being in this field, is that what we can learn from using technologies to, to kind of send surveys to people or using kind of passive monitoring, it's not just that we can then develop interventions. But I think we can also use what we learn from these technologies to feed back into our theories. So classically, in health psychology, our theories have been developed based on observations that were made kind of time points that were quite spread out. So for example, a baseline assessment, and then four weeks later, we might have looked at how something related to an outcome. And then building a theory around, for example, how people respond to health threats, or sort of how people's intentions or attitudes are linked to behaviors. But actually, our kind of current theories are static. So they can't really account for sort of how people evolve over time in context, which we know from research using technologies to kind of better understand what happens to people is that people change over time. So our theories aren't really kind of suited to, or well, I think, kind of the, the harshest critique we can give of these current theories is that they're not actually capturing kind of what's happening in in the real world.


Lili Golmohammadi  17:18

And the temporal changes that come with these technologies that we have in our pockets? I was reading a little bit about behavior change and apparently, the original theories they were developed in the - was it the 70s, maybe 70s?


Olga Perski  17:31

Yeah, some of them, yeah. 


Lili Golmohammadi  17:33

Yeah, yeah, and they still been kind of they've continued. And I think it's only from now what you've been saying as well, it's only now that people are kind of re-engaging with those and rethinking those theories that have kind of underpinned research for many decades.


Olga Perski  17:46

Yeah. And even though there are moves towards kind of incorporating what we know about how people change over time, or the the temporal changes, it's still quite a challenging process, given that health psychology is also a very applied field. And it's important that what we learn can then be implemented into healthcare, for example. So that translation process is probably also going to take quite a long time. So the work that we're doing now is probably not necessarily going to be applied into healthcare for another couple of decades. So I also think that's why kind of some of these theories from the 70s and 80s, that most health psychologists and people in digital health know are probably not very useful, they're still being used a lot kind of in practice.


Theme music  18:38


Lili Golmohammadi  18:46

So how has collaborative work you've been doing with computer scientists developed your research skills and approach?


Olga Perski  18:53

Yeah, so I think working with people who are experts in quite different disciplines, I think it kind of makes you think really carefully about sort of what skills I have, or what would be useful skills in order to kind of take me to where I want to be or to do the work that I want to do. And I think in these collaborations, I think I've particularly become a lot more interested in kind of computational modeling, and also things around machine learning, but kind of the computational aspect, which then requires a bit more MAPS knowledge and also coding skills. So those are kind of some of the areas that I'm working on at the moment to kind of try to improve my skills in because I think that will then kind of help support the work and it doesn't necessarily mean that I have to execute everything myself, but I think it's really important for kind of understanding where I'd like to get to or kind of where the direction the projects could go in if you do have kind of more in depth knowledge of some of these skills.


Lili Golmohammadi  20:03

Yeah. So the type of evidence and methods that you use in health psychology and computer science in human computer interaction, as we've been discussing, they're quite different, have you experienced any challenges or tensions trying to work across them? And how have you managed these?


Olga Perski  20:21

This type of interdisciplinary working does come with some quite important challenges, not to say I mean, there are, of course, so many benefits and that's why this work is so important. But I think it can be quite challenging coming up against language barriers, for example. So we might be using different terms, but actually, we kind of talk about very similar things. But then also, I think, a very important barrier is kind of having misaligned goals and kind of expectations of the nature of the work and the nature of the collaboration.


Theme music  20:58


Lili Golmohammadi  21:06

I was hoping that you could offer some advice for other social researchers who wants to do collaborative research with computer scientists or other disciplines?


Olga Perski  21:14

Yeah. So I think there's quite a few things that could be done that I'm actually very excited about, because I think this is the future. And I think it's kind of well, first of all, I think what people tend to mention when or at least what I've noticed is kind of people in my field, for example, tend to talk about that it's about kind of finding the right people to collaborate with. So you kind of want to find somebody who understands you and your discipline, and then kind of stick to that person. And I think that's perhaps a bit sort of misconstrued in that, it shouldn't necessarily have to happen that way. Because I think that actually displays a bit of well, that there aren't necessarily kind of enough resources put into it or enough value placed on being a bit interdisciplinary. So I think being the type of researcher who can kind of bridge different disciplines is really valuable. So in my field of digital health, I kind of think of this as an inherently interdisciplinary field. And there are now some master’s programs being established where all of these different traditions are all come together and where they draw on methods from different disciplines. I think that's also incredibly valuable that kind of there will be a new generation of researchers who already know these different traditions for kind of people who are already quite experienced researchers, I think, kind of encouraging people to take courses or just widen their horizons a bit, I think is also incredibly useful for making sure that people do have a bit of a better understanding of other disciplines and don't kind of get tunnel vision and just focus on their own.


Lili Golmohammadi  22:58

Thank you Olga so much for your time.


Olga Perski  23:00

Well, thank you so much for having me. It was really, really nice to meet you all.


Lili Golmohammadi  23:08

You've been listening to Together We Create. This episode was presented by myself Lili Golmohammadi and edited by Cerys Bradley. I was joined today by Olga Perski. If you want to find out more about her research or the podcast series, please follow the links in the description. This podcast is brought to you by the UCL Collaborative Social science Domain