Why do some people wear a Fitbit?

I was thinking back to a conference that took place some time ago on the theme of data capture, in particular to the presentations on wearable physical activity devices or trackers [1]. These were still fairly new at the time and I quickly picked up that the people in the audience, most of whom were much younger than me, were quite sniffy about them. This contrasted with their enthusiasm for hearing more about music streaming. We talked about this over coffee and one participant mentioned his disappointment that his partner would only listen to the same songs again and again on Spotify rather than accessing their Daily Mix (a collection designed to extend your listening based on what you had already accessed). I sympathised. In contrast there was not much talk about activity trackers; they were not cool and they were not going to take off.

Well, I don’t think trackers have turned out to be cool but they have slowly taken off, at least among certain social groups, though nothing like to the same degree as social media [2].  But it remains easier to see the limitations of the wearable devices than their advantages [3]. An obvious constraint on take-up is that users have to invest in hardware (wearable devices) to get the full benefit and this costs money and is another piece of kit you need to look after. There are doubts too about the accuracy of trackers and concerns over privacy and sharing of data.  Activity loggers can play into people’s anxiety about health and advertising can take the form of  gender stereotyping.  Ceaseless logging of physical activity can become an end in itself so that you end up gaming the system, e.g choosing your terrain carefully to increase step count. For that matter you can cheat outright – for example by attaching the device to a bike or dog! The key point, however, is that if you are comfortable about your fitness level the devices serve no real purpose; if you are naturally fairly active you just get on with it, you don’t need to measure what you are doing. This is even more the case when it comes to logging sleep activity. Those who sleep well don’t talk about it, record it or even think about it, they just do it.  In fact were they to monitor sleep they might well become more self-conscious and disrupt what was working perfectly for them in the first place.

So why trackers? They can help users who have a special need to focus on physical activity and this seems to be particularly the case as you get older and are aware of becoming less active. The evidence is fairly slim, surprisingly these are still early days in research of activity logging [4], but as Ridgers et al. (2016) put it ‘there are some preliminary data to suggest these devices may have the potential to increase activity levels through self-monitoring and goal setting in the short term’. This is not a ringing endorsement but sounds about right to me.

Aside from large scale quantitative work, we also need to know more about the experiences of wearing these devices. Again there are some papers often based on what particular types of wearer get out of it, e.g. those recovering from serious illness, using activity as part of weight reduction programme, older people. The key point made by Jarrahi et al (2018) is an obvious one, but worth repeating: if you are disposed to see wearable devices as motivating then you will find them to be, if not forget it.

Noticing gaps in the literature I spoke to people I knew who used activity trackers and asked them for their opinions. They had found the devices generally useful as they helped with focusing on activity levels. Users had particular goals in terms of improving fitness or weight watching so that if they noticed their step count was falling they would deliberately do something about it, i.e. go for a run or walk. Some spoke of friendly competition with others. The devices were worthwhile though their usefulness was linked to the short term goals they had set themselves, future use was less clear.

I don’t wear an activity tracker. However, I did once have a Garmin watch which I used for a while for tracking runs. I got some satisfaction from knowing I was improving my speed and I liked to check that I was pacing myself evenly. I got out of the habit of using the device when, as age was catching up on me, my times were getting longer rather than shorter. Of course logging might be most useful when performance is dropping but I felt knowledge of my own tail-off was only go to depress me. I also found charging the device to be a faff and worried about the power running out. At one point I was disappointed about a run because the battery had discharged half-way through and I would not get a full reading to download. This was ridiculous. Why did I need to measure it to believe I had done it and enjoyed it? However, even if I no longer use my watch, I still line up with other runners after a weekly Park Run to get my time recorded [5]. We clearly have no need to do this, so why do we do it? I found some comments by Engeström [6] useful here. Drawing on Vygotsky and Russian social constructivism, Engeström sees exercising agency (i.e. getting to do what we would know we want to do or must) as a two fold process: first design it and then do it (the execution phase). He gave the example of an alarm clock. Finding the will to get up earlier is not easy. The alarm clock helps reminds us that we must do it. However, an alarm set by someone else would not work, we need to have planned for its use, i.e. we calculate what time we need to get up and set the alarm accordingly. In fact very often the planning is enough and we wake up any way before or as the alarm goes off. I think tools to measure physical activity work in a similar way. We design the use of the tool, i.e. we personalise the device with our details, we look at captured data and we make judgments about what we need to do, and having designed it we feel encouraged to do it. We are using the device to put into effect an otherwise vague intention to take more exercise; the device is helping us do it. I don’t want to go out and buy a tracker or Smart watch but I am not as dismissive of them as my fellow conference participants once were. And I will still queue to log my time at Park Run even if the data are only pointing in one direction.

[1] Wearable devices will capture data on movement, including how far one has moved and how many steps completed. They will also do a calorie used count and may monitor heart rate, sleep length and sleep activity. Waterproof devices will do distance and strokes when swimming. Data can be shared by users.  Probably the best known companies for producing wearable devices are Fitbit, Garmin, Huawei and Withings. A smart watch is not dedicated to physical activity but will include physical activity tracking. A smart phone App such as Map My Fitness can log runs,  but does less than a wearable device.

[2] Estimates of numbers of users are in the tens of millions, e.g. 30 million estimated active Fitbit users and 55 millions Apple watch wearers, 20 million Map My Fitness users, These are large numbers but not in the same league as, say, numbers of Facebook users.

[3] There are several contributions on trackers and other devices to ‘The Conversation’ which are fairly sceptical of their value, e.g.:

Siek, K. (2020) Why fitness trackers may not give you all the ‘credit’ you hoped for [online]

https://theconversation.com/why-fitness-trackers-may-not-give-you-all-the-credit-you-hoped-for-128585

Duus, R. and Cooray, M. (2015) How we discovered the dark side of wearable fitness trackers

https://theconversation.com/how-we-discovered-the-dark-side-of-wearable-fitness-trackers-43363

Kerner, C., Quennerstedt, C. and Goodyear, V. (2017) Young people oppose Fitbits in schools [online]

https://theconversation.com/young-people-oppose-fitbits-in-schools-84311

[4] There are several systematic reviews, e.g. Ridgers et al. (2016) and Shin et al. (2019), with most concluding that there is not much to systematically review in the first place.

[5] Park run is a free 5km run held in many parks in UK [https://www.parkrun.org.uk] and now around the world.

[6] Engeström’s key example concerns ‘cheating slips’ used by students. These are notes which student might access during an exam but Engeström argues it is the making of the notes rather than the access to them that make them effective.

References

Engeström, Y. (2006). Development, movement and agency: Breaking away into mycorrhizae activities. In K. Yamazumi (Ed.)  Building Activity Theory In Practice: Toward The Next Generation. Osaka: Center for Human Activity Theory, Kansai University. (CHAT Technical Reports #1).

Jarrahi, M., Gafinowitz, N. & Shin, G. (2018) Activity trackers, prior motivation, and perceived informational and motivational affordances. Personal and Ubiquitous Computing, 22, 433–448.

Ridgers, N., McNarry, M. and Mackintosh, K. (2016) Feasibility and effectiveness of using wearable activity trackers in youth: a systematic review.  JMIR mhealth and uHealth 4, 4: e129.

Shin, G., Jarrahi, M.H., Fei, Y., Karami, A., Gafinowitz, N., Byun, A. and Lu, X. (2019) Wearable activity trackers, accuracy, adoption, acceptance and health impact: a systematic literature review. Journal of Biomedical Informatics, 93: p.103153.

 

 

 

 

Big Data and Education

Reading through contributions to debates on big data and education (see [1] – [9] below) reminds us that there are two perspectives on education: education as science and education as wisdom of practice.

Education as science is about generalisation (if x what is likely to happen?), measurable outcomes (chiefly grades, if appropriate retention too) and observable processes of teaching and learning. Education as science seeks ‘objectivity’ and appeals in particular to policy makers and organisations who want to know what is worth funding and what works.

Education as wisdom of practice is about recognising complexity. From experience teachers know that no two classes are ever the same and without wisdom gained over time they will lack the strategies and techniques to get the best out of their learners. Wise teachers recognise uncertainty both in what is being measured and how you measure it. They ask how and why people learn, or fail to learn, and build up mental pictures of their learners. They know that teaching and learning is emotional. The appeal to wisdom of practice plays very well with practitioners who understand that ‘it is always complicated’.

Big data (and learner analytics) appeal instinctively to our first group, those who see education as a science. This on the grounds that:

  • We can interrogate more data, more quickly, than ever before. For example in England the national pupil database (NPD) [1] provides a huge data file on pupil attainment, attendance and progression which can be broken down by sector, gender ethnicity, first language, free school meals and so on. It can help show what is happening and what is having an impact. To take one example it has helped us understand not just about ethnicity, but different dimensions to ethnicity, and it has recalibrated how we think about impact of multi culturalism in schools and identified an achievement gap among white working class boys.
  • We can be serendipitous. Indeed other than its size the NPD is not big data, it is a snapshot survey with predictable data fields. Of more interest to big data enthusiasts are explorations of hitherto unlinked sets of data. For example projects in several universities aim to explore what data on library usage (not just visits, but what is accessed online) have on the attainment of students [2] [3]. It is not, of course, surprising if library usage and attainment are in some way related but we have never been able to explore the association in as much detail as now.
  • We can get just in time data which can be fed back to learners, for example to ‘nudge’ students to visit the library [2] [4]. This kind of early feedback can be very important in distance learning where a host of online activity is automatically tracked. [4]
  • We can use big data in order to generate powerful models which assist decision-making at policy, institution and teacher levels. Such modelling has enabled the construction of so called dashboards for learners allowing them to compare their activity against both the performance of the group and ideal profiles [2]. Of course this monitoring has come under scrutiny for overgeneralisation. More nuanced commentators have noted that big data analysis throws up association rather than proof of causality [6] It is also essential for researchers to consider the ethics of it all [3]. Indeed ethical issues extend beyond respecting confidentiality, and awareness of the data protection legislation, to wider questions such as whether we really want to live in a society in which data on our activity is automatically generated. [7]

The Big Data movement is not immediately attractive for our second group, those who believe in the wisdom of practice. In particular:

  • Those using Big Data often ignore that teachers collect data all the time both formally and informally. For example teachers know which assignments seem to engage students and which do not, and they can adjust when they see pupils bored and unhappy. Above all, they have the back story about the students they teach which means that they can interpret actions sensitively and intelligently [8]. Practitioners bring an understanding of context and develop compelling metaphors for thinking and learning which inform and explain practice.
  • Big data may be big but it does measure everything. Rather it is only measuring what can be measured not what it is appropriate to measure. One, perhaps unintended, consequence is that outcome measures are dominating thinking about education but this is swamping debates about educational values, for example what should be taught and how we should assess it.
  • Those obsessing about Big Data simply miss the fact that teaching is a complex mix of different types of knowledge. Insights from data analysis often confirm what teachers have long thought and often articulated among themselves. They are left asking why no-one spoke to them in the first place.

Is there a way forward between teaching as science and teaching as art? Probably not, but:

  • First we can understand that different stakeholders might be asking different questions about data and it is quite understandable that they do so. For example one statistical association that many institutions have found is that those attending a library induction will often achieve better outcomes than those miss the induction. This is useful for managers wanting to make decisions about resourcing libraries and for course leaders in planning their courses and promoting such induction. What of course it does not tell us is how to teach a class nor should we conclude that if you attend library induction you will get a better degree. Big data might not provide a story about learning but that is fine as long as we recognise the limitations.
  • Second, we can use data as one more source of information about teaching and learning. Indeed as teachers we can misread the signs and cues and we can become highly focussed on what we do, on our performance, rather than what the learners do, both in and out of formal teaching sessions. In practice teacher performance may turn out to be less important than we think, it is the design for teaching that matters more and this is something for which Big Data can help us [9].
  • Third, we can take the Big Data movement with a pinch of salt but can accept the wider case that we should make more use of available data to inform teaching. Indeed most practitioners do understand that so-called objective data about learner activity can help, especially when explored as part of practitioner research projects. What we need is, perhaps, a Small Data revolution, in which local data is used to support practitioner inquiry before we talk about a Big Data movement.

Update

This post was in anticipation of a one day event on Big Data at the University of Warwick.  At the event Ben Williamson presented a more disturbing view of learning analytics or at least subjected the claims made by private educational providers such as Pearson to critical scrutiny. He also noted the scale of the funding and ambition of commercial organisations to provide comprehensive profiling of learners, going far beyond the rather tentative examples of higher education. For more go to https://codeactsineducation.wordpress.com/2015/12/09/education-big-data-imaginary/

References

[1] For more on the National Pupil Database in England go to: https://www.gov.uk/guidance/national-pupil-database-apply-for-a-data-extract

[2] A JISC report gives a useful snapshot of learning analytics in higher education: in Sclater, N. (2014) The Current State of Play in UK Higher and Further Education published at:http://repository.jisc.ac.uk/5657/1/Learning_analytics_report.pdf

[3] A parallel report looks at legal and ethical issues: Sclater, N. (2014) Code of Practice for Learner Analytics, at http://analytics.jiscinvolve.org/wp/2014/12/04/jisc-releases-report-on-ethical-and-legal-challenges-of-learning-analytics/

[4] Opportunities for formative use of data are presented in Ferguson, R. and Buckingham Shum, S. (2012). Social learning analytics: five approaches. In: 2nd International Conference on Learning Analytics & Knowledge, 29 Apr – 02 May 2012, Vancouver, British Columbia, Canada, pp. 23–33.

[5] A feel for learning analytics in respect to online activity is given in, for example, Fidalgo-Blanco, A. et al (2015) Using Learning Analytics to improve teamwork assessment, Computers in Human Behavior, 47 (2015) 149–156 though this is not specific to DL.

[6] An argument for understanding complexity is put by Beer, C. et al (2012) Analytics and complexity: Learning and leading for the future, presented at ASCILITE2012 Future challenges, sustainable futures. http://www.ascilite.org/conferences/Wellington12/2012/images/custom/beer,colin_-_analytics_and.pdf

[7] At the time of writing there is some consternation among librarians in Japan that a newspaper breached confidentiality by publishing the names of books that novelist Haruki Murakami, 66, took out as a teenager from his school library, this all seems rather innocent: http://www.theguardian.com/books/2015/dec/02/librarians-in-uproar-after-borrowing-record-of-haruki-murakami-is-leaked

[8] This is argued robustly by Buckingham Shum, S.  Learning analytics: white rabbits and silver bullets, University of Technology, Sydney in Williamson, B. (ed.) 2015. Coding/Learning: Software and digital data in education. Stirling: University of Stirling at: https://codeactsineducation.files.wordpress.com/2015/08/coding_learning_-_software_and_digital_data_in_education.pdf

[9] For an exploration of learning as in this data rich world try Goodyear, P. (2015) Teaching as Design in (ed P. Kandlbinder) A Review of Higher Education Vol. 2. www.herdsa.org.au/publications/journals/ herdsa-review-higher-education-vol-2

Pessimistic narratives about technology (continued)

‘They’ are collecting data on us every day, who we phone, where we move, what we buy, who we see, what we do and say. Some at least of this is benign; if ‘they’ know more about where we go then we might be able to have a more rational transport system; if they know what we want there is a better chance that we will get it. But what if ‘they’ have more sinister motives. This was the theme of the ‘Dictatorship of Data’ [1] in which the BBC correspondent Gordon Corera looked at the use of big data in surveillance societies. There were several key themes in the programme:

• The personal data routinely collected today far exceed what was collected in even the most obsessive surveillance societies such as DDR or imagined in dystopian novels such as Orwell’s 1984.
• We are right to think that our actions and movements are being monitored and even if they are not, we increasingly imagine that they are with all the consequences that brings.
• You cannot escape from being monitored. For example the programme spoke to an Ethiopian dissident who found asylum in England only to discover that his movements were being monitored via his laptop by Ethiopian security services, putting any one he contacted at risk.
• There is a flourishing and uncontrolled trade in surveillance. For example there are commercial organisations [2] willing to provide surveillance services to nearly anyone for anything.
• You do not need to monitor everything. For example it might be more useful to access your list of social network ‘friends’ than to know what you are actually talking about.
• Social media, so often seen as the tool for opening up new forms of counter cultural protest [3], provide unexpected opportunities for security agencies to harvest lists of dissenters and to manipulate and disrupt through rogue messaging.

Two further issues which came out of the programme had a more general significance in how we think about Big Data. First, analysis of Big Data only works as our lives are patterned and fairly predictable. We might like to think of ourselves as spontaneous and creative but in practice we are not; we need regularity and order in our relationships and because of this we are traceable. Second, the sheer quantity of Big Data might appear overwhelming and to search for dissenters might look like searching for needles in a hay stack. However with Big Data the ‘haystack’ provide the clarity. In other words deviations from the norm stand out because the norm is so clearly established.

Programmes on the perils of Big Data can easily get stuck into dystopian views of technology but Gordon Corera largely avoided this by offering counter cases. For example he gave space to speakers from the Tactical Technology Collective (TTC) [4] an organisation concerned with ethical use of Big Data in the service of social change. However, as with nearly all reporting of technology, it was difficult to avoid a narrative of inevitability regarding both the impact of technology and our responses to technology. In practice technology has always had unpredictable consequences, those or who predict the future often get it wrong. As an example there were voices in the 1970s which proclaimed that the introduction of technology would mean shorter working weeks and unimagined opportunities for leisure, but compare this to what actually happened [5]. Part of this unpredictability, and something that social science can never resolve to everyone’s satisfaction, is how do we recognise both order / pattern and change / agency. In many discussions of big data [6] there is an ‘ecological fallacy’ which leads researchers to extrapolate from noticing patterns of group behaviour to the assumption that anyone who shares certain characteristics of that group will behave in a similar way. What is more, there is a backdrop to our behaviour which requires explanation: circumstances change and people change with them. As Corera’s programme showed, in spite of unprecedented surveillance, the DDR collapsed and at some time point in time so will the present Ethiopian government. Finally, the programme left you wanting to know more about Big Data and ‘liberal democracy’. Rather than a binary distinction between bad and good regimes doing bad and good surveillance there is surely a continuum.

[1] BBC (2015) The Dictatorship of Data, 17 November 2015 Radio 4 available at
http://www.bbc.co.uk/programmes/b06pb831
(Note that programmes are usually available for a limited time only).
[2] The programme spoke to FinFisher representatives – for more on Finnischer go to https://en.wikipedia.org/wiki/FinFisher. You would guess, given their willingness to speak, FinFisher were by no means the worst example in this murky field.
[3] Castells offers one of the most romantic perspective here: Castells, M. (2012) Networks of Outrage and Hope: Social movements in the internet age, Cambridge: Polity Press.
[4] The Tactical Technology Collective is at https://tacticaltech.org
[5] Robins, K. and Webster F. (1988) Athens without slaves…or slaves without Athens? The neurosis of technology. Science as Culture 1: 7-53.
[6] A similar point is made in a growing critical literature on Big Data, see for example Kitchin. R. (2013) Big data and human geography: Opportunities, challenges and risks, Dialogues in Human Geography, 3, 3. 262-267. For a chattier article see
Cukier, K. and Mayer-Schönberger, V. (2103) The Dictatorship of Data, MIT Technology Review May 31, 2013 at www.technologyreview.com/news/514591/the-dictatorship-of-data/

Big Data

There is a lot of interest, and indeed funding, around Big Data. Big Data is a catchy – and suitably vague – phrase which is used to draw attention to the increasingly large amounts of data available to policy makers, natural scientists, and social researchers. Definitions of Big Data are still up for grabs but many centre on the three V’s: velocity, volume and variety. Going further some stress that Big Data are generated in networked systems so that data can be shared across contexts and updated in ‘real time’. One further aspect of Big Data is that while we can set up systems to get at the data we need there is also a lot of data generated for other purposes, or indeed with no clear purpose in mind. Discussion of Big Data often slips over into the Internet of Things – another vague term but one that points out the myriad ways in which we are getting constant streams of data through sensors, monitoring devices and tracking of network activity.

The concept of Big Data is naturally being pushed by hardware and software companies, but it is getting a sympathetic following in government too. It is a big idea in academia and is a priority for research funding [1].

We can be cynical about the concept of Big Data but there is a genuine fascination with the way that technology is changing how we investigate and control our environment. Many of the examples of the application of Big Data seem benign enough – for example monitoring water levels so that early warnings can be given about floods or drought; waste bins that are fitted with sensors so that councils know which bins need emptying and when; car movements tracked so that planners can get immediate feedback on changes to road layouts. However people start getting worried about other contexts. For example music analytics allow the industry to spot emerging musicians and genres but may do so at the expensive of originality; political parties are increasingly adept at tailoring their appeal to different constituencies based on data about those they are targeting; our online movements are tracked in order to build up profiles of us as consumers. Each example of data collection might not be sinister in itself but taken together contribute to our increasing sense of being manipulated. Going further, data are, every day, gathered in ways which most of us would find intrusive if we knew how much was being collected and by whom. A small scale example of this is provided by this blogger (Siraj Datoo) talking about sensors fitted to recycling bins:

http://qz.com/112873/this-recycling-bin-is-following-you/

So what stand should we take on Big Data? One perspective is to realise that many of the issues to do with Big Data are not new – we know how to raise questions about privacy, access rights, security, use and misuse and we need to raise these again. We should also be aware of past attempts to harness large sets of data. For example a forerunner of Big Data was so-called ‘freakonomics’, Levitt and Dubner [2], which crossed over from academia into the book buying public. Freakonomics explored associations within large sets of data and in some cases played around with the data to see what came up – one well publicised example was Levitt and Donohue [2] who claimed an association between legalised abortion (since 1973) in USA and a drop in crime 18 years later. A lot of the examples put forward about Freakonomics were interesting but there was always a question mark, as there is today, about drawing conclusions from ‘associations’ between data rather than offering an explanation or an indication of causality; in other words just because there appears to be a relationship between A and B (they go up or down together) does not mean that A causes B or vice versa. It is ourselves that do the interpretation of the data and the decisions based on data are political ones – for example the monitoring of water levels is in itself very useful but monitoring will not make the water appear or tell you who gets the water and who does not.

We need reminding not to suspend our professional expertise, still less our common sense, when faced with masses of data or (or more correctly algorithms which have organised the data into what seem meaningful ways). A well reported case here is airline pilots’ overreliance on auto pilot and a supposed difficulty in taking decisions when faced with real life emergencies; a more everyday example would be as a motorist when you follow the Sat Nav even when it is taking you north and you want to go south or when you blindly follow directions into a flooded road. Going back to our earlier example a political party or music label will only get so far following trends, no matter how up to the minute their monitoring is, they also have to have a feel for the enterprise they are engaged in and in some way seek to set the agenda.

If looking to be optimistic about Big Data try the NGO Ushahida: http://www.ushahidi.com

For a sympathetic review go to: http://www.ssireview.org/articles/entry/open_source_for_humanitarian_action

Or try the United Nations’ organisation Global Pulse. http://www.unglobalpulse.org/research

I am sure both have their critics but they really are helpful for seeing that technology and political and social action are not incompatible.

[1] The Big Data Family is born – David Willetts MP announces the ESRC Big Data Network http://www.esrc.ac.uk/news-and-events/press-releases/28673/the-big-data-family-is-born-david-willetts-mp-announces-the-esrc-big-data-network.aspx

[2] Levitt, S. and Dubner, S. (2005) Freakonomics: A rogue economist explores the hidden side of everything, New York: William Morrow/HarperCollins.

[3] Donohue, J. and Levitt, S. (2001) ‘The impact of legalized abortion on crime’, The Quarterly Journal of Economics, 116, 2, 379 – 420.