User:Leighblackall/Data and Power



An argument that 'privacy' is not the right word to frame the current issues around data. The word doesn't connect with the unsettled issues we have with corporate and state collection and use of it. The issue is more accurately about power, and through this frame we connect to a range of critical, political and economic ideas that are more helpful to consider.

Log

 * March 2018 Drafting a book chapter
 * 20 July 2017 Connecting An ethical framework for ubiquitous learning, as a Learning Analytics CoP forms at RMIT and to include in a presentation to eLearning Korea 2017.
 * 16 December 2016 After Joyce Seitzinger mentioned Paul Prinsloo's work interrogating there ethics of data, I had a brief exchange with Paul on his blog post Decolonising the collection, analyses and use of student data: A tentative exploration/proposal then through Twitter, Paul offered a link to Seeing Without Knowing: Limitations of the transparency ideal and its application to algorithmic accountability Mike Ananny, Kate Crawford First Published December 13, 2016, notes to follow...


 * 29 July 2016 After finally subscribing to the Corbett report, a journalistic video cast by James Corbett in Japan, I became aware that the TOR Project is quite likely spyware. I unfortunately recommended the TOR project as a basis for defense against data retention, at 18:50 in the recording of the talk. For more details, see James Corbett and Pearse Redmond's 2013 discussion


 * 4 October 2013 Alex Hayes and I briefly discuss co authoring a paper with the focus of 'data analytics for people learning, not of people learning'.


 * 1 October 2013 Video published on Youtube, Archive and Wikimedia Commons


 * 26-27 September 2013 Presented to University Analytics Forum. As with all the presentations I do, these are initial ideas and thoughts that will hopefully generate comments to guide a more focused look. If it goes anywhere, I'll keep notes on a Wikiversity page


 * 2010 Contribution to the Learning Analytics email group discussion

Leigh Blackall 2017 Copy on Google Docs Work in progress

Keywords: networked learning, connectivism, algorithmic learning, quantified self, collectivist society, learning analytics, big data, artificial intelligence

Introduction
I'm sitting in a dark empty house, training YouTube's algorithm to make better music recommendations through their new “Songs” application. I’m trying to work out what actions I can take to elicit the most interesting recommendations. At the same time I’m wondering, what are the various biases built into the software? How far into indy music can it reach? What different discovery methods can I bring into it? Can I maintain more than one account? What interests are being served? How might algorithms like this be used to assist personal learning? If only the documentation on how it worked was publicly available, there must be thousands of people wanting to discuss these questions more thoroughly.

Teaching Youtube to Teach
In 2016/17 I was involved in a not dissimilar project that was trying and work out how to “teach youtube to teach”. A small group within a School of Fashion and Textiles at RMIT University were exploring the possibilities of learning through the links and recommendations made by Youtube’s algorithms. We first spent some months working with a teacher to set up a Youtube channel into which we created and uploaded instructional videos, created playlists and subscribed to a range of other channels. We then asked the students to come to a computer lab so we could observe the effects of certain simple activities together. We asked the students to bring up Youtube and note the recommendations being made before creating an account. We then asked them to create a new account and return to the Youtube home page and see if anything in the recommendations and search seemed different. Then we asked them to find the teacher’s channel and subscribe to it and and again notice the differences. We showed them how to reject recommendations, and accept others, consciously customising their new Youtube account into specific subject areas and notably improving the recommendations and search results that YouTube was making. By extension then, the same methods could be applied in Facebook and Google search. We were observing changes in the recommendations after completing a simple list of activities.


 * 1) Teacher creates a channel
 * 2) Students create a new account or channel and subscribe to the teacher’s channel
 * 3) Teacher and students respond to Youtube’s recommendations
 * 4) Teacher and students create playlists
 * 5) Teachers and students subscribe to each other
 * 6) Teacher and students network outwards by subscribing, playlisting, liking and responding to other content across Youtube.

Yes, we did notice marked improvement to the search results, recommendations, and associations being made by the algorithm, and we wondered what could be achieved with more time and with a larger network.

Can we teach machines to teach
After the Youtube experiment, I decided to test similar methods in the relatively recent platform called ResearchGate. After setting up a profile, making some connections and giving some responses to the things the platform was recommending to me, I posed a question, “Can we teach the machine to teach - using algorithms as a learning resource?” Quite quickly responses from other ResearchGate users started coming in, from people I was not already connected to. I found myself in quite satisfactory discussion with people who had interesting perspectives and greater experience to offer. Because the connections were entirely new, I can only assume that the platform’s algorithm was facilitating the connections by putting my question in front of enough people with similar interests. The interactions were short lived however, despite my best efforts to honour the responses people offered, and trying to sustain collaborative inquiry. It seems the platform did not successfully alert those responders to my replies, or my replies did not draw out further participation from any of them.

I did gain a useful reading list from the brief interactions that I would not have easily obtained through search.

Mager, R.F. 1961. What are teaching machines doing to teaching. Audio Visual Communication Review. Vol. 9, No. 6 (1961), pp. 300-305 - It is fascinating to read articles like this, from relatively so long ago, writing about a technology in such tones as though the disruptions are imminent and far reaching. Reminds me so much of today. I especially enjoyed the summary, where it called out the newly created professional role of the AV Specialist, who had already apparently missed the significant technologies of the time - language learning and TV teaching. This rebuke reminded me of the thorough demolition of professional roles by Ivan Illich in his 1978 book, "The Disabling Professions" (Illich 1978), a phenomenon I continue to observe in my recent history experience of university staff completely misunderstanding Wikipedia and Youtube, or the IT staff not comprehending the the significance of Skype and Hangouts before investing millions in lecture capture tethered to theatre AV… etc

I was troubled by Mager neglecting to offer a clear picture of what the "self learning [and/or] teaching machine" was thought to be in 1961? Without saying so, Mager made it clear enough that he was imagining some sort of content delivery platform, based in clear "behavioral goals" and "subject matter organisation". I was interested to see this emphasis on content, despite being written in an era when more connected learning platforms were already being envisioned by the likes of Vennavar Bush As We May Think (1945) through to Douglas Engelbart Mother of All Demos (1969).

I think this distinction between content driven learning vs connection driven learning might be crucial. I experience the conceptual tension almost daily while working in universities that primarily fund and resource outcomes-based, subject organised, structured content development work such as websites and interactive media (textbooks), but have little to no capacity to comment on the self evident learning phenomena throughout the internet, based in informal teaching and learning and connectedness.

I first wrote about this repetitive disconnect between formal and informal, professional and ameture, authorative and folk, in 2005 with Early Film, Early Internet, Early Days, Networked Learning. That post connected me to Gerry White who was heading up Education Australia, and invited me to present, Teaching is Dead, Long Live Learning at Global Summit 2006. The year before Gerry had written Beyond Horseless Carriages (White 2005), drawing out the annunciation of Stephen Downes’ tour of Australia the year before that.

These are machine facilitated connections, made through the data we each gave to the machine. Had we not, we may not have connected and engaged our interests in such a way, for several years hence. The same sort of connection can be replicated, in Youtube, ResearchGate, and others. How it can be sustained and checked remains an interesting problem.

Bring Your Own Account
In 2015/16 I was working with the ever-engaged and creative thinker Frank Buechele, on a small and self directed project to make the notion of “Bring Your Own Account” tenable at RMIT University. We named the initiative that way to try and ride the relative success of an earlier transformation in the university - “Bring Your Own Device”. The idea was simple: people should be able to sign in to the university’s systems using an already existing account that they probably have outside the university, and in doing so connect a wider range of data relating to before, during and after their enrolment. In many ways the proposal had a lot in common with the arguments made 10 years earlier, that people should be able to bring their own computer devices and have them simply work on the network. One of the key value propositions this time though was the opportunity for the university to collect and analyse data from a wider spectrum of activity than the narrowness of its internal systems within an extremely limited time and space of engagement.

Our principle inclination in this though, was for a “student facing data” conception of learning analytics. {ref} To try and devise a system that could be used by people to better access the data about themselves, manage their data and, if they want, compare it to others in a de-identified pool. We hoped at least some people would discover useful insights for themselves by agreeing to connect their accounts and data. We wanted a way for people to connect and disconnect their data at their leisure, and to be motivated by clear and well considered reasons for doing so. We saw other conceptions of teacher and manager dashboards as inherently problematic.

Similar concepts were emerging, like the FitBit for Education analogy of learning analytics, and research questions about teaching machines to teach. On that basis we looked at combinations of available applications, from time management, project management, self improvement and browser profiles. Conceivably it seemed possible to combine these apps, encourage their use in assignments and activities, and invite exploration of their usefulness by looking at resulting data.

These concepts carried the principle of a more distributed and interconnected use of data, and thinking this way lead us to realise the ground work already done by people in the quantified self movement. Here was a relatively long history of people exploring the use of a variety of devices and methods to generate data on themselves and optionally pooling their data with the sets of others to potentially gain insights on themselves, and in relation to others. {ref} That practice seemed entirely relevant to the notion of Learning Analytics, and a much richer, more ethical approach than other conceptions being developed.

Quantified self for learning analytics
In Defining Networked Learning (Blackall 2014) I described a method for quantifying learning outcomes from a timeline of my own distributed and networked learning data. While it was primarily associated to the ideas of informal learning webs (Illich 1971, Brown 2008) and situated participatory learning (Lave and Wenger 1991), it also happens to be an example of “Quantified Self” - a mostly computing and internet subculture where people generally engage in the generation and collection of data about themselves, to analyse for insight. My method was to recount a narrative of informal learning and link in available data to help build out a narrative of learning with evidence of outcome.

This method of learner quantified self was published in Alex Hayes’ paper to the 2013 IEEE International Symposium on Technology and Society (ISTAS) Identity awareness and re-use of research data in Veillance and social computing (Hayes 2013). Hayes used the example among others to propose that data generation and publishing is itself an act of research with value in the academic publishing tradition. Likewise, I was arguing that data can be used to validate informal learning toward assessment practices within formal education.

Quantified Self came to broader attention in 2007/8, as the large scale effect of socially networked media was becoming clearer. (Hesse 2008, Wolf 2010). Of course there were threads to this movement reaching back well before 2007. An ontology approximately encompassing transhumanism, sousveillance and open data for example, and these domains connect to an even broader arenas of collectivism and connectivism.

A quantified self approach to learning analytics carries within it principles of user agency and control. Such principles might be distilled to the concern of power, and through such consideration, different and perhaps better formed ideas for learning analytics can be made.

Lupton categorised five modes of self-tracking can be identified. These include private (for one’s own purposes only); communal (sharing data with other self-trackers); pushed (encouraged by others); imposed (foisted upon people); and exploited (where people’s personal data are repurposed for the use of others). (Lupton 2014)

Privacy and Power
Where data is collected and used for the automated observation and intervention on people’s options, behavior and preferences, “privacy” is a common keyword used to represent a range of concerns and considerations. In terms of Learning Analytics, there has been calls to discuss issues that are broader than privacy, such as a critical appreciation of the ideological underpinnings of the power holders in data collection for actionable learning analytics (Slade 2013).

It might be useful to consider for example, the confidence universities and software developers have when approaching learning analytics through their collection and retention of student and teacher activity online, and whether or not that same enthusiasm would sustain if the equivalent developments we're taking place in the physical spaces of learning. Would they see no problem tracking student and teacher attendance at lectures and tutorials; or their presence on campus and in specific venues like libraries; or their participation in social events; or to sample conversations via their phones, in a bid to improve recommendations and inform other well-meaning interventions? Most would reasonably expect a definite ‘no!’ to such ideas, or would they? In the online sidepaces there is barely a concern, despite the almost complete dissolve of any clear distinction between online and off (Lupton 2014). It can be thought provoking in most instances, to take what has become common practice in online teaching and learning generally, and to imagine an equivalent implementation and level of acceptance in more traditional teaching and learning spaces like classrooms, labs, studios and internships. The keyword Privacy seems to assume a certain level of common understanding as to its meaning and implications, and has become a dominant ethical consideration in the popular conception around the governance of data and related development, at the expense perhaps of other pressing issues.

When we turn our attention to the idea that data is being used to make recommendations to people, and that those recommendations focus, limit or restrict people’s exposure to information; and if we were then to consider how that could direct people’s knowledge and behaviour en mass, then we’re talking about power, a sort of power that Shoshana Zuboff has termed, Surveillance Capitalism (Zuboff 2015).

AgitProp memes abound that point out the apparent imbalances of power around data, and these are gradually feeding into what we must consider with learning analytics. For example the ~2016 polemic from The Free Thought Project that sets up Julian Assange and Wikileaks vs Mark Zuckerberg and Facebook: The text with the two portraits reads, “Hi. I’m Julian Assange. I give private information on corporations and government to you for free and the media calls me a criminal. Hi, I’m Mark Zuckerberg. I give your private information to corporations and government for money and the media called me man of the year.”

The independent journalist James Corbett takes such polemic further with a review of now open propaganda techniques used by governments and military through what Corbett calls, The Weaponisation of Social Media (Corbett 2018). “Now openly admitted, governments and militaries around the world employ armies of keyboard warriors to spread propaganda and disrupt their online opposition. Their goal? To shape public discourse around global events in a way favourable to their standing military and geopolitical objectives. Their method? The Weaponization of Social Media”

Power then, is a useful keyword to represent a range of concerns and considerations relating to data, analytics and software. It acknowledges a certain level of inevitability around the loss of privacy and introduces the consideration of who gains, who loses, and how. The emphasis on privacy may have contributed to an obfuscation of questions relating to power in the popular discourse, which now plays out in the polemics above, and probably should be brought down into the development circles around learning analytics.

By proposing that we consider data to have issues with power, I mainly hope to demonstrate how a basic change in keyword can afford a different perspective on the domain and, with that perspective, different types of invention and development might be made possible. In the area of learning analytics, the “student facing data” principle is an example of an approach derived from the consideration of power in data. Lupton uses the term “data double” to describe the use of data to generate a depiction of people that they choose to engage and act on, or not (Lupton 2014 P6).

The questions become, how can we create systems that give people the most control over the data they are generating, and the most benefit? Where can the balance be found between that and the potential for institutions to better respond to people’s needs, and how can those institutions avoid falling into the well known trap of serving their own interests over the people they are supposed to serve? (Illich 1978) How can institutions be mindful of the diverse and sometimes divergent interests of the individual people they serve? And how can institutions be more mindful of the interests that their “industry partners” bring to the domain as well? These and more questions abound when taking power into consideration of data and analytics.

Collectivism and Connectivism
The internet’s technological foundations in the strategic military concerns of mutually assured destruction, combined with 60s and 70s new-age counterculture entrepreneurialism that was largely reacting to the military madness (Dammbeck 2003, Curtis 2011, Herzog 2016) and radically reimagining the spiritual/political/social/economic arrangement, has developed a digital platform steeped in approximate ideologies and functionally reorganising society through the sheer scale of the technological influence. A richly networked economy (Benkler 2006), a self organised everybody (Shirky 2008), even mob wisdom! (Surowiecki 2004). A new form of freedom (Stallman 2010).

A few voices have tried to confront what they see as utopianism in and around the technology, pointing out its inherent ideological nature (Bowers 2000), or the darker side of humanity inside its gates (Keen 2007), or the how alone we seem to be in this hyper connected world (Turkle 2012). More recently a new and successful genre storytelling has emerged called Tech Paranoia, brought to us in the surprising form of the television series Black Mirror.