Sea Hero Quest: using play data for dementia research
25 July, 2017
Sea Hero Quest (SHQ) is a free game available on Android and iOS, which collects data on how players navigate its wayfinding tasks and then uses this data as part of dementia research. Amazingly, it has translated 73 years' worth of gaming activity from 2.7 million players into the 'largest dementia study in history', and so I was excited to speak to Dr Hugo Spiers (Reader in Cognitive Neuroscience, University College London Institute of Behavioural Neuroscience), who is leading this research.
Creating a diagnostic tool for Alzheimer's dementia
The game tasks the player with navigating in a boat using simple left/right controls. Data from players will show how well they use spatial navigation, and on a larger scale will show how spatial navigation ability differs across various demographics, including across ages.
With this vast bank of general population data, the team are taking the first step in developing a tool to detect signs of dementia, by understanding what normal spatial navigation behaviour is, and what is pathological.
The initial levels of SHQ feel more like tutorials. These levels serve a purpose within the study, explains Dr Spiers: 'We're not tapping into the brain areas that you need to find your way... you are just going to a visible target and that allows us to effectively measure how fluent you are with these types of video games.'
A seasoned player would reach the target faster than a gaming novice. This allows the team to understand how experience with gaming impacts on the player's navigation skill, and to account for this in their analysis.
The team can use mathematical modelling to assume everyone is bad at videogames, 'and you still see a decline... just the steepness of the slope is much shallower because they start off performing badly in their early age group; they're already poor navigators, but it's exactly the same relationship between age and performance, it's just the angle of the slope has changed because your starting point is different.'
There are plans to trial the game with an orienteering team, to see how their performance in the game relates to real-world performance. 'It might show that there are certain levels that seemingly, for some reason or another, allow you to predict the best navigation performance in the real world. Or certain bits of those levels in breaking them down.' This will help the team in honing a more precise diagnostic tool.
The team have already presented preliminary findings; perhaps most notable is that spatial navigation ability as captured by the game appears to deteriorate with age from the early 20s onwards. These findings were based on two month's analysis of a set of levels that tested the player's ability to know which direction they had travelled from.
There are more findings to come as the team sets to work analysing data on how the players actually navigated around the wayfinding levels. 'As soon as you go to the tracking in the main wayfinding levels, we have a 500 millisecond resolution of where people were and what direction they were facing,' Dr Spiers notes. 'It allows you to look at their choices in the topological network of paths; it allows you to look at how much they are deviating from the optimal path.'
Yes, there are people in the general population who happen to be bad navigators. But the wayfinding levels should give enough navigation data to tease out specific differences in navigation ability which may indicate deterioration due to dementia.
'Something we have from the clinical literature is that people with dementia are more likely to just keep going forward than to take a left or right turn,' says Dr Spiers. 'It might be that you have healthy bad navigators who just make lots of bad choices. But then you maybe have someone who's got early stage Alzheimer's dementia, and they literally just go forward into the wall and they don't deviate in their path... the way they move is very different, diagnostically.'
Dr Spiers notes that most cognitive exams for dementia that have a lot of precision are language-based, which makes it hard to equate them across countries. 'Plus, the clinicians I speak to aren't really convinced that Alzheimer's dementia is a verbal memory problem... patients are actually quite good at verbal memory, early on in the disease. But they get lost: that's why they typically come in to the clinic.'
Gaming as a research tool
I had made it to level 40 of the game by the time I had spoken to Dr Spiers. With around 30 minutes of game time clocked, the game told me I had contributed two days' worth of research data to the project.
This was quite astonishing to me. Dr Spiers explained how this was the case. 'We optimised the game in a way we hadn't done before to collect this really detailed trajectory data, so you couldn't normally use the trajectory at that level, every 500 milliseconds. It wouldn't be that meaningful in a small sample, but because you're contributing to a very large sample it means every 500 milliseconds we're getting very useful information from you. But if you were coming to the lab... to do this in the lab setting is just ridiculous; it's not possible.'
Dr Spiers considered how other fields of mental health research could benefit from this form of study. Movement monitoring may prove useful for people with recurrent depression.
For example, someone with severe depression is less likely to leave their house. But before they reach that stage, there may be more subtle changes that could be picked up. "The patterns in which they move through the space around their community before they stop leaving home might be predictive of a future onset of severe depression. So there's potential in data on that which would be interesting to look at.'
While SHQ isn't looking at players' real-life movements, the team is learning a lot about how they can analyse trajectories, which is the bulk of data they get from the game. "People vary enormously in what routes they take and it's quite possible we could explore how people with depression change the way they move within the game.'
When it came to designing the game, Dr Spiers found the team having to compromise between gaming and science. He's pleased the game has received generally positive reviews on the app stores. 'I thought if we'd not achieved that, it'd be a disaster because the fun game component does relate back to the science: if it's not a fun game, you're not going to get 2 million people playing it.'
There were also compromises made in the game design in order to fit the needs of the experiment. 'If it had been a raw experiment, if someone was coming into my lab, they would have spent five minutes filling out a load of demographic questions. If you'd started the game online with five minutes of demographic questions, our sample size would be a thousand, because nobody would want to sit and answer a load of personal questions about their life before they get into the game, it doesn't work like that on the app store.'
Instead, the game asks only nine demographic questions about the player. 'As a scientist, we needed to really work hard to get the best questions into the game. But we always knew it had to be done carefully, as a compromise, asking 9 rather than 90 questions.'
Dr Spiers foresees a rise in the use of games in research and gamifying experimental data input. 'It is really good... It's not typically seen to be worth it, whereas every experiment I make in my lab, I try to gamify it, so it is engaging, because I believe that you want people at their optimally engaged for your experiment... There's some experiments where I think it's just not going to be worth it. But I think in things that we're interested in (navigating environments), you do want to bring people up to a situation that they're familiar with: which is generally gaming.'
With millions of people playing video games in their free time, there is clearly great potential here for gathering data on playing activity for cognitive behavioural research, from players who agree to participate in such research. As SHQ has shown, the data acquired can be huge and detailed, which can provide new insights into population behaviour. It also lowers the barrier for people to take part in research, making it easy for people to participate from their own gaming devices.
With the study involving SHQ already showing fascinating preliminary findings ahead of deeper data analysis, the field of gamified behavioural research seems to hold a very exciting future.
Sea Hero Quest is part of a research project that is a collaboration between Alzheimer's Research UK, University College London, University of East Anglia, and the UK game development team Glitchers.
Authored by Sachin Shah