In Education, the bots patrol the data lake in the hope that ‘datafication’ of students, reconstructing people from data (Knox et. Al. 2019, Enyon 2015), will lead to penetrating insights about how they learn. The promises of AI and data are being used uncritically with a view to personalising learning (Tsai and Gašević 2020, Bulger 2016). The process seems to me to flow from the existence of the data lake itself, the promise of ‘datafication’ or perhaps quantification to ‘learnification’ (Knox et. al. 2019) and then to personalised learning (Ibid.). Learnification is the turn toward a more student-centered approach aimed at allowing the learner to choose her own road-map through the datascape. Combining the two developments only opens the doorway to personalised learning (Knox et. al. 2019) through a series of well-intentioned nudges and operant conditioning (Ibid.).
The ideal of personalised learning is not itself new, as Socrates himself often taught one-to-one, the golden ratio in Education (Friesen 2019). Friesen (2019) argues this ideal is imagined and unobtainable, essentially it cannot exist in Higher Education. Yet I feel Frank Smith offers some hope when he points out the intimate role of reading when it is linked with the power of learning identities (Smith 1998:23). In other words, motivated students can seek their own masters once clear of the white noise of information in Higher Education.
In my visualisations this week I have examined this white noise in the form of information I receive about my students’ learning and I also wondered just how personalised this is. And finally, I measured my own one-to-one’s with written educational material. In my working week, software that even promises personalisation does not yet exist, but I find the one-to-one’s with my students offer real personalisation and reading offers me something similar. Data technology has some way to go on this score. But it would be wrong to suggest data is not important. Edwin Hutchins, an anthropologist, observed US Navy navigation teams conducting navigational computation, he concluded the Navy personnel formed one part of a cybernetic system of humans, machines and data. He concluded the learning process is as follows:
Building a structure based on the availability of data
Describing the organisation of knowledge
Dividing up knowledge by dividing computational tasks between people (modularising)
Fitting the computational and social spaces
Hutchins (1995: 324)
This is somewhat like the process of data > datafication > learnification>personalisation (Knox et. al. 2019). Furthermore, I suggest the personalisation of learning is largely a matter of identity as suggested by Smith (1998). This identity is, of course, also a matter of social fit. Inevitably, machine learning attempts to mimic an established human behaviour. But since we know very little of how humans think AI has not moved from the Chinese Room argument, simulation not assimilation.
Bibliography
Bulger, M. 2016. Personalized Learning: The Conversations We’re Not Having. Data & Society working paper.
Eynon, R. (2015) The quantified self for learning: critical questions for education, Learning, Media and Technology, 40:4, 407-411.
Friesen, N. 2019. “The technological imaginary in education, or: Myth and enlightenment in ‘Personalised Learning.” In M. Stocchetti (Ed.), The digital age and its discontents. University of Helsinki Press.
Hutchins, E., 1995. Cognition in the Wild (No. 1995). MIT press.
Knox, J, Williamson, B & Bayne, S 2019, ‘Machine behaviourism: Future visions of “learnification” and “datafication” across humans and digital technologies‘, Learning, Media and Technology, 45(1), pp. 1-15.
Smith, F., 1998. The book of learning and forgetting. Teachers College Press.
Tsai, Y-S. Perrotta, C. & Gašević, D. 2020. Empowering learners with personalised learning approaches? Agency, equity and transparency in the context of learning analytics, Assessment & Evaluation in Higher Education, 45:4, 554-567.
When we read we can also join the company of authors. We can share ideas and experience with them…
Frank Smith, “The book of learning and forgetting” (1998: 24)
In 1450 Johannes Gutenberg invented the printing press, making it possible for ideas and experience to be shared, for the first time, globally. Knowledge could be the preserve of anyone who could read, rather than the few who owned hand written manuscripts. This did more to democratise learning than any modern learning technology.
This week I recorded how much learning I did through the medium of reading. I included the reading I am doing for the MSc in Digital Education (I am currently doing two courses together), reading for work and finally, in my leisure time. Technically, the latter category qualifies as I rarely read fiction, but if I did I would still be learning language, characterisation, story-telling etc. MSc reading is just when I am reading articles and book chapters. Work reading is usually the same kind of material, I am not counting e-mails etc., just journal articles, text books etc.
Frank Smith (1998:25) likens reading to apprenticeships, since you learn at the feet of the aithor, this is dialogue, this is true learning. Yet we increasingly denigrate it as not sophisticated enough, preferring instead, perhaps, a video presentation, passive “banking” of knowledge (Freire 1993:45).
This week was, admittedly, a good week I was reading for 4 to 5 hours throughout the week, I am proud I managed to do this around teaching, marking and admin. My apprenticeship continues.
References
Freire, P., 1996. Pedagogy of the oppressed (revised). New York: Continuum.
Smith, F., 1998. The book of learning and forgetting. Teachers College Press.
Activity tracking is a lot more straightfoward than than measuring learning. The number of clicks on lecture videos or reading activities tells one nothing about what is going on in their heads. On the other hand, regular running is known to improve health, so if my activity tracker says I went for a 5K run each day, then I can be fairly sure I am keeping relatively fit. And if I need further reassurance, I can check my heart rate which is effortlessly measured constantly, and is known to be a good measure of fitness. My resting heart rate is just above 50 and when I was ill recently it shot up to 90 for days, showing just how responsive it is as a measure of stress.. So it is reliable and simple. This just goes to show that although we can be rightly critical of narrow quantitive measures, they can be extremely useful when we know what we are measuring. Of course, I could argue that I did not really need to buy a running watch to know that I run regularly and that this is good for me. The popularity of fitness trackers and the quantification of our lives is, ironically, much harder to understand. In the times before activity trackers or, perhaps, before jogging became a thing, people were much fitter than they are today.
This week I wanted to get away from a very linear approach to data visualisation by creating an image which encapsulated a more networked approach in an attempt to capture the messier nature of data flows. I fear the image is somewhat messier than I hoped. But perhaps this serves to illustrate the point. I followed the approach used by David McCandless in using images that convey the subject matter and altering the size according to the magnitude of each variable (2012). In this case I wanted to measure, over the week, just how personalised or student-centered learning analytics data were. Reading through the articles on personalised learning (Tsai and Gašević 2020, Friesen 2019, Thompson and Cook 2017, Bulger 2016) I was struck by that although these articles engage with a critical debate about personalised learning, the data at my disposal does not claim to create individualised learning assessments. I am currently teaching 109 post-graduate business students using a simulation game. The simulation platform (Cesim) collects data about time on the platform as does the accompanying Moodle site. The students also communicate on MS Teams. Additionally I receive emails and talk to students on Zoom. My analysis shows how each of these platforms are connected and uses a traffic light system to indicate how personalised these platforms are. Unsurprisingly, the face to face Zoom calls are the only opportunity to provide personalised feedback.
References
Bulger, M. 2016. Personalized Learning: The Conversations We’re Not Having. Data & Society working paper
Friesen, N. 2019. “The technological imaginary in education, or: Myth and enlightenment in ‘Personalised Learning.” In M. Stocchetti (Ed.), The digital age and its discontents. University of Helsinki Press.
McCandless, D., 2012. Information is beautiful (pp. 978-0007294664). London: Collins.
Thompson, G. and Cook, I. 2017. The logic of data-sense: thinking through learning personalisation. Discourse: Studies in the Cultural Politics of Education. 38(5), pp. 740-754
Tsai, Y-S. Perrotta, C. & Gašević, D. 2020. Empowering learners with personalised learning approaches? Agency, equity and transparency in the context of learning analytics, Assessment & Evaluation in Higher Education, 45:4, 554-567
Learning with data is perhaps how we imagine the future will be. Perhaps a future where the teacher has been replaced completely by data, a landscape of knowledge where the teacher no longer shows the way (Biesta 2012). Perhaps in this landscape bots and web crawlers will assist the mythical self-directed learner to find what she needs. With 44 Zetabytes of data on the internet, much of it on social media this might be a good place to start if one wanted to pursue the dream of open and free learning. With this in mind, I considered four days’ worth of potential learning on social media for this week’s data visualisation. Since this week has included a Tweetorial, I thought this would be a good opportunity to incorporate this in my visualisation. I confess I am a reluctant Twitter user but I feel it is necessary to keep up with the more informal information around Higher Education, which I usually hoover up passively. I spend more time on LinkedIn and this is because academia is my second career, and in industry LinkedIn is more popular for professional learning and networking. I am yet to shake the habit. And finally, I included WhatsApp, which, again is used widely in business as a free way to coordinate and communicate across the Globe. Incidentally, many of my business friends are migrating to Signal and Telegraph in light of new data privacy arrangements at WhatsApp. I might have included Instagram but I only post pictures of sunsets and my family there, I haven’t looked at Facebook for several years.
I wrote down each time I used social media this week, categorising the data by platform, and whether I was learning directly, indirectly, purely socialising or networking, Finally I took note of which apps ‘nudged me in certain directions, one way which platforms might use one’s data (Tsai and Gasevic 2020, Knox 2019). These nudges could be towards useful data, which might assist me as a learner, or to commercial sites, or towards similarly minded people, potential friends. The data were hand drawn using the icons to record 10 minutes worth of social media time on each platform, categorised as explained above, with the extra notation of nudges received on that day by platform.
Data visualisation decoded
I have recreated the visualisation above with the key to decode the categories typed out, as my handwriting is so bad. In terms of direct learning, I spent an hour on Twitter on three of the four days. This is unusual actually as I do not use Twitter that often normally, perhaps a few minutes a day. However the Tweetorial was extremely educational and perhaps this is something I should do more often. Previously, I have used LinkedIn for this purpose but I would be learning what is happening in the world of business. Since I lecture at a business school, this makes sense, the 20 minutes I spent learning about business initiatives on Monday is more like my daily routine. Indirect learning is how I have labelled information I receive which indicates where to find learning resources, how to access something, or perhaps where I might find a good course on a certain subject for free, the logistics of learning, if you like. The majority of these data were from WhatsApp as I am a member of the MSc (Digital Education) group, established on IDEL. I spend a fair amount of time on there both helping and being helped. What little socialising I do on the internet is split between all the platforms. This highlights to me how little I enjoy the social aspect of these platforms. I would probably spend less time doing this if were not for the isolation of covid-19. I see social media as a resource, one to which I give and receive data, knowledge and even advice. This probably has a social aspect to it, but I find the purely social interaction awkward. Finally, I use social media for professional networking, mostly LinkedIn. Again I see this as making myself available to help people (mostly) and occasionally receiving support and advice. This is arguably social learning and altruistic collaboration, perhaps a far older method of learning that the dialectic method discussed by Friesen (2019).
I also noted the nudges in terms of recommended content. This was found only on Twitter and LinkedIn, it was generally commerical in nature and somewhat incompetent in actually making recommendations that suited me. We actually discussed this in the Tweetorial, the hype around AI would suggest we should be getting far more accurate nudges by now? I experience these nudges as noise. A constant tugging away from one’s purpose to follow this person or to a commercial website. Although social media is not expressly a learning platform, it is seen by many as having this capability. But a learning environment, for me, is quiet. Like a library. Both LinkedIn and Twitter are noisy places, there are many distractions. LinkedIn has a learning function, called LinkedIn learning, which I have used in the past. While I sat through endless videos, notifications streamed in, constantly urging me away from my purpose. I guess it is because learning from others (social learning) is actually quite noisy, there are gems of knowledge amongst this noise.
This might be an argument for the internet as a learning environment, that here we may learn from others or even bots in an imagined one to one environment. That the library and classroom are unatural constructs which sever us from social data in favour of decontextualised data. Human cognition, afterall, is believed to be built around social intelligence (Chudek at al 2012). In 1885 a man called Ebbinhaus demonstrated chopping up information into tiny decontextualised chunks made learning very difficult and forgetting very likely. The forefathers (for they were men) of the modern education system thought this was an excellent idea and built our education system on exactly the method Ebbinghaus had shown to be so problematic (Smith 1998). The information we tend to remember is usually tied up with social significance which is why the apprentice model of learning has been historically popular. But the one-to-one model has proved, unsurprisingly, to be ineffective for mass education (Friesen 2019).
So we hope to recreate the one-to-one scenario with electronic tutors and guides for our new digital landscape, even if this is a fantasy (Friesen 2019). The nudges are programmed using Bernard Skinner’s pidgeon box logic and behavioural economics are as ubiquitous on social media as they are in the supermarket (Friesen 2019, Knox 2019). Based on the rules of reinforcement or operant conditioning, we are offered treats for behaving correctly. Yet the AI is patently not up to the task, sending us on random errands, based on something we read last week, for example. Despite break throughs like the GPT-3 natural language algorithm of 2020, we remain in an AI Winter. Consequently, nudges are noise,not digital Platos. Perhaps one day AI will break the paradox of using Big Data to understand individual choice and autonomy. Whilst bots may predict behaviour on average, they cannot predict the behaviour of individuals. These mutually incompatible levels of analysis are what make personalised learning chimeric (Tsai and Gasevic 2020, Knox 2019, Bulger 2016). Nudges are noise.
Biesta, G. J. J. (2012). Giving Teaching back to education: responding to the disappearance of the teacher. Phenomenology and Practice, 6(2), 35–49. https://doi.org/10.1074/jbc.275.7.4949
Bulger, M. (2016). Personalized Learning: The Conversations We’re Not Having. Data & Society, 1–29. Retrieved from https://datasociety.net/pubs/ecl/PersonalizedLearning_primer_2016.pdf
Chudek, M., Heller, S., Birch, S., & Henrich, J. (2012). Prestige-biased cultural learning: bystander’s differential attention to potential models influences children’s learning. Evolution and Human Behavior, 33(1), 46–56. https://doi.org/10.1016/j.evolhumbehav.2011.05.005
Friesen, N. (2020). The Technological Imaginary in Education: Myth and Enlightenment in ‘Personalized Learning’. In Stocchetti M. (Ed.), The Digital Age and Its Discontents: Critical Reflections in Education (pp. 141-160). Helsinki University Press. doi:10.2307/j.ctv16c9hdw.12
Knox, J., Williamson, B., & Bayne, S. (2020). Machine behaviourism: future visions of ‘learnification’ and ‘datafication’ across humans and digital technologies. Learning, Media and Technology, 45(1), 31–45. https://doi.org/10.1080/17439884.2019.1623251
Smith, F., 1998. The book of learning and forgetting. Teachers College Press.
Tsai, Y. S., Perrotta, C., & Gašević, D. (2020). Empowering learners with personalised learning approaches? Agency, equity and transparency in the context of learning analytics. Assessment and Evaluation in Higher Education, 45(4), 554–567. https://doi.org/10.1080/02602938.2019.1676396
I am an academic working in a business school. There are two things I really enjoy: teaching and research about teaching (which is called scholarship). And like most people, I really don’t like admin. Is it because I am sensitive to admin that I think I am always doing it and that it distracts from teaching? I suppose you might say there is teeny bias at play here.
As a try-out data visualisation, I thought I would record my emails for a week to see how many are, in fact, admin e-mails and whether my hunch that I am doing way too much is actually true. Collecting data is very easy, just rummage through my inbox for a weeks emails. As it happens I colour code my emails as admin, scholarship and teaching. I have a system. I split my working day roughly three ways giving a certain amount of time for each category. I bias my hours towards teaching because I figure that’s mainly what I get paid for, I also really love teaching. I started to do this when I noticed that admin was taking over my working day.
So this fairly straightforward data collection is ideal for a first go at data visualisation using hand-drawn methods. In the week I chose (week 2 2021), I received 244 emails, of which 177 were admin, 53 teaching and 14 were scholarship. Ha ha! as I thought! I am still spending too much time on admin. Hand drawing the data gave me time to really think about the data itself. I was initially reminded of a time before digital, of hand-written notes and also memos and letters. In 1989 I joined the Royal Navy as an officer. On basic training, as well as the usual running around, we had to learn how to write various letters, such as memos, routine letters, official and demi-official letters. Official letters were signed:
I have the honour to be,
Sir,
Your Obedient Servant
Can you imagine that? You couldn’t make mistakes on official letters, but in other kinds you could cross out the mistake neatly and then write it correctly. In the digital world we just edit, but then it took too long to start again every time. On the postcard, you can see I made a mistake on the total 244, which I wrote as 144, and then just wrote over it. I really couldn’t start again.
The visualisation shows an overwhelming number of admin emails, I could barely fit them in. The other types of email look pitiful. So job done?
Williams (2017:29) points out that the data we select should really be called capta because it is only the data we select not the whole data set that we observe. Selwyn and Gasevic (2020) discuss the limitations of data in learning analytics. The fact is, in learning analytics, and other data analysis we cannot escape our biases, something the analysts themselves freely admit (Ibid.). The limitations of these data are that all emails are not equal in terms of th time theye take to process or action, some are read and deleted, some are deleted and not read, others are the start of a huge project. There is no way from my analysis to tell which are for information and which for action. I also do a lot of teaching communication on MS Teams, running courses is an operational task and Teams is better suited for this than email. In short the visualisation has dramatic effect but the devil is in the data selection.
