01 June 2016

Can we teach the machine to teach?

Analog Computing Machine By NASA on Wikimedia Commons


When Michael Wesch published the short video The Machine is (Using) Us in 2007, he was summarising a long discussion about digitisation and hypertext, search engine algorithms and optimisation (SEO), The Singularity, The Semantic Web, Crowd Source, Open Source, Peer to Peer (P2P), Artificial Intelligence (AI) and many other ideas all within four and a half minutes.

The core question Wesch was asking was whether these machines are using us, or if they are us. Most people I talk to about this tend to focus on the idea that the machines are simply enticing us to their front end web services, so as to collect data, our media, our interactions, our interests, our responses, our location and movements.. to create powerful demographic and individualised profiles for marketing and other intelligence gathering purposes. But perhaps the machines are also offering us a reflection, using the collected data to then make recommendations and associations that can be both surprising and useful and also meaningful in a reflecting sense - showing us what they make of us.

Some people go further than Wesch's questioning, and ask if we could in fact use these machines more deliberately, if we could teach these machines to connect and teach us? Could we input data and information more consciously, to get the sorts of recommendations and associations we need? Would the machines start offering us even more unexpected and useful connections? Might our research improve, might our networks widen and grow in value? Would our knowledge and productivity exponentially increase?

Take the simple action of using a hashtag for example (or tag more generally). These people-generated metadata identifiers are a way to connect with potentially vast repositories of similar information, are information in and of themselves, and perhaps more importantly - connect to people. The potential of this was first captured to video in 2005 by Jon Udel looking at Delicious (an early and very popular social bookmarking service).

About here is where the theme folksonomy started to emerge as a serious alternative to taxonomy. Related to this was the idea that flat, non-hierarchies and comparatively anarchic ways of organising online, were proving successful.But tagging (and hypertext more broadly) is significant but simple technology in comparison to our current questions around digital identity management. And the relatively recent webservice called Pinterest offers a tangible experience of machine connections and associations. It users socially bookmarked images, text and weblinks to make its connections and recommendations, and as almost anyone who uses the service will testify, with impressive effect. Certainly the field of SEO and social media marketing have long been interested in these ideas, but are teachers, students and researchers?

When I started the Wikipedia entry for Networked Learning, I was interested in the people-to-people and people-to-content connections that the socially networked Internet was facilitating. Later, I wanted to look into the plausible history of Networked Learning and discovered a range of related ideas from as early as the industrial revolution - when railways and telegraph lines were being erected across vast distances, through to post industrial ideas in the 1970s such as Ivan Illich (Deschooling Society) and Christopher Alexander et al (A Pattern Language) - who were seminal in imagining a more dis-intermediated, post industrial arrangement for teaching, learning and education.
The operation of a peer-matching network would be simple. The user would identify himself by name and address and describe the activity for which he sought a peer. A computer would send him back the names and addresses of all those who had inserted the same description. It is amazing that such a simple utility has never been used on a broad scale for publicly valued activity.” Ivan Illich (1971). Deschooling Society - Chapter 6
Now we have an Internet that is more socially connected, might we explore Networked Learning for ideas on how to make better connections, associations and recommendations for teaching and learning?

So the questions for me become: can we teach the machines to teach us? Are there certain activities and methods, projects and assignments, identities and roles we can use that will help us input more exacting and impact-full data so the machines will make the sorts of recommendations and associations we need to advance our access to information and people, and ultimately our knowledge and understanding? Can I use commercially orientated machines like Youtube in such a way that it will start recommending more and more useful videos, and start connecting me to more and more valuable people? Can I do the same with Facebook? Can I do the same across Google? Can it be done through Wikipedia? Can I do this differently with multiple identities? Or will marketing, mass news, propaganda and advertising prevail? Will xenophobia and surveillance disrupt and distort the potential?

I've been experimenting with these ideas for 10 years now. In 2013 I attempted to quantify some of this with a research project called Defining Networked Learning, where I analysed a trail of data from my own efforts to self learn a reasonably technical body of knowledge and skills relating to biomass heating. This work was published in 2014 through the IEEE in a paper called Identity Awareness and Re-use of Research Data in Veillance and Social Computing.

Others have done much better work, looking at human to human interactions of networked learning before the Internet. Work such as Entienne Wenger and Jean Lave's well known Communities of Practice and Legitimate Peripheral Participation. Such work was used by people in the Education sector to expand and develop ideas around Student Mobility and Life Long Learning. To my knowledge, few if any have updated these ideas with experience of a socially networked Internet and the deliberate use of it to generate connections and associations.

We're looking at this form of machine teaching and learning, in various places around the College of Design and Social Context at RMIT.

  1. We've actively challenged the University's current practice of offering Google accounts to staff and students and then suspending those accounts when a staff member or student is no longer active with RMIT.
  2. We've coined the tag "Bring your own account" working with people in ITS to discuss ideas around what it might look like if the university was to accept and use a person's preferred account (online identity) when they join, and to shape that account in the time they work with us. These sorts of ideas have big implications for things like Learning Analytics (currently focused narrowly on data gathered in an issued account within an extremely limited timescale).
  3. We advise teachers in the Schools to help students to think about the consequences of building an RMIT issued account - only to have it suspended. Implied in this advice is to encourage people to bring their own accounts and shape them instead, so what they build can go with them, after they graduate (or conversely - what they've already built comes in, and is shaped over the time they use the account with us).
Now we're looking for ways to develop, quantify and measure this approach to online learning. Our current ideas are to look at both the RMIT issued accounts as well as the own accounts brought in. We're looking for how we might quantify these accounts over time, as their users undertake activities specifically designed to teach the machines to teach - to build rich online identities so that professionally relevant recommendations and associations are being made by the 'machines' by the time a student graduates, and hopefully able to take their account/s with them.

If you're interested in working with us on these questions, or to simply join in discussions relating to them, please make contact. Leave a comment or email us directly. There's potentially exciting work to be done.

No comments: