On the importance of place

Posted July 4, 2024 by Jon Dron

I had the great pleasure of being invited to the Open University of the Netherlands and, later in the day, to EdLab, Maastricht University a few weeks ago, giving a slightly different talk in each place based on some of the main themes in my most recent book, How Education Works. Although I adapted my slides a little for each audience, with different titles and a few different slides adjusted to the contexts, I could probably have used either presentation interchangeably. In fact, I could as easily have used the slides from my SITE keynote on which both were quite closely based (which is why I am not sharing them here). As well as most of the same slides, I used some of the same words, many of the same examples, and several of the same anecdotes. For the most part, this was essentially the same presentation given twice. Except, of course, it really, really wasn’t. In fact, the two events could barely have been more different, and what everyone (including me) learned was significantly different in each session.

This is highly self-referential. One of the big points of the book is that it only ever makes sense to consider the entire orchestration, including the roles that learners play in making sense of it all the many components of the assembly, designed for the purpose and otherwise. The slides, structure, and content did provide the theme and a certain amount of hardness, but what we (collectively) did with them led to two very different learning experiences. They shared some components and purposes, just as a car, a truck, and a bicycle share some of the same components and purposes, but the assemblies and orchestrations were quite different, leading to very different outcomes. Some of the variation was planned in advance, including an hour of conversation at the end of each presentation and a structure that encouraged dialogue at various points along the way: these were as much workshops as presentations. However, much of the variance occurred not due to any planning but because of the locations themselves. One of the rooms was a well-appointed conventional lecture theatre, the other an airy space with grouped tables, and with huge windows looking out on a busy and attractive campus. In the lecture theatre I essentially gave a lecture: the interactive parts were very much staged, and I had to devise ways to make them work. In the airy room, I had a conversation and had to devise ways to maintain some structure to the process, that was delightfully disrupted by the occasional passing road train and the very tangible lives of others going on outside, as well as an innately more intimate and conversational atmosphere enabled (not entailed) by the layout. Other parts of the context mattered too: the time of day, the temperature, the different needs and interests of the audience, the fact that one occurred in the midst of planning for a major annual event, and so on. All of this had a big effect on how I and others behaved, and on what and how people learned. From one perspective, in both talks, I was sculpting the available affordances and constraints to achieve my intended ends but, from another equally valid point of view, I was being sculpted by them. The creators and maintainers of the rooms and I were teaching partners, coparticipants in the learning process. Pedagogically, and despite the various things I did to assemble the missing parts in each, they were significantly different learning technologies.

The complexity of distance teaching

Train journeys are great contexts for uninterrupted reflection (trains teach too) so, sitting on the train on my journey back the next day, I began to reflect on what all of this means for my usual teaching practice, and made some notes on which this post is based (notebooks teach, too). I am a distance educator by trade and, as a rule, with exceptions for work-based learning, practicums, co-ops, placements, and a few other limited contexts, distance educators rarely even acknowledge that students occupy a physical space, let alone do we adapt to it. We might sometimes encourage students to use things in their environments as part of a learning activity, but we rarely change our teaching on the fly as a result of the differences between those environments. As I have previously observed, the problem is exacerbated by the illusion that online systems are environments (in the sense of being providers of the context in which we learn) and that we believe we can observe what happens in them. They are not, and we cannot. They are parts of the learners’ own environments, and all we can (ethically) observe are interactions with our designed systems, not the behaviour of the learners within the spaces that they occupy. It is as hard for students to understand our context as it is for us to understand theirs, and that matters too. It makes it trickier to model ways of thinking and approaches to problem solving, for example, if the teacher occupies a different context.

This matters little for some of the harder elements of the teaching process. Information provision, resource design, planning, and at least some forms of assessment and feedback are at least as easy to do at a distance as not. We can certainly do those and make a point of doing them well, thereby providing a little counterbalance. However, facilitation, role modelling, guidance, supporting motivation, fostering networks, monitoring of learning, responsive adaptation, and many other significant teaching roles are more complex to perform because of how little is known about learning activities within an environment. As Peter Goodyear has put it, matter matters. The more that the designated teacher can understand that, the more effective they can be in helping learners to succeed.

Because we are not so able to adapt our teaching to the context, distance learning (more accurately, distance teaching) mostly works because students are the most important teachers, and the pedagogies they add to the raw materials we provide do most of the heavy lifting. Given some shared resources and guided interactions, they are the ones who perform most of the kinds of orchestration and assembly that I added to my two talks in the Netherlands; they are the ones who both adapt and adapt to their spaces for learning. Those better able to do this in the first place tend to do better in the long run, regardless of subject interest or innate ability. This is reflected in the results. In my faculty and on average, more than 95% of our graduate students – who have already proven themselves to be successful learners and so are better able to teach themselves – succeed on any given course, in the sense of reaching the end and achieving a passing grade. 70% of our undergraduates, on the other hand, are the first in their family to take a degree. Many have taken years or even decades out of formal education, and many had poor experiences in school. On average, therefore, they typically have fewer skills in teaching themselves in an academic context (which is a big thing to learn about in and of itself) and we are not able to adapt our teaching to what we cannot perceive, so we are of little assistance either. Without the shared physical context, we can only guess and anticipate when and where they might be learning, and we seldom have the faintest idea how it occurs, save through sparse digital signals that they leave in discussion forums or submitted assignments, or coarse statistics based on web page views. In a few undergraduate core courses within my faculty it is therefore no surprise that the success rates are less than 30%, and (on average) only about half of all our students are successful, with rates that improve dramatically in more senior level courses. The vast majority of those who get to the end pass. Most who don’t succeed drop out. It doesn’t take many core courses with success rates of 30% to eliminate nearly 95% of students by the end of a program.

Teaching with a context

We can better deal with this if we let go of the illusion that we can be in control and, at the same time, find better ways to stay close: to make the learning process including the environment in which it occurs, as visible as possible. It is emphatically not about capturing digital traces and using analytics to reveal patterns. Though such techniques can have a place in helping to build a picture of how learners are responding to our deliberate acts of teaching, they are not even close to a solution for understanding learners in context. Most learning analytics and adaptive systems are McNamara Machines, blind to most of what matters. There’s a huge risk that we start by measuring the easily measurable then wind up not just ignoring but implicitly denying that the things we cannot measure are important. Yes, it might help us to help students who are going to get to the end anyway to get better grades, but it tells us very little about (for instance) how they are learning, what obstacles they face, or how we could help them orchestrate their learning in the contexts in which they live. Could generative AI help with that? I think it might. In conversation, an AI agent could ask leading questions, could recommend things to do with the space, could aggregate and report back on how and where students seem to be learning. Unlike traditional adaptive systems, generative AI can play an active discovery role and make broader connections that have not been scripted. However, this is not and should not be a substitute for an actual teacher: rather, it should mediate between humans, amplifying and feeding back, not guiding or informing.

For the most part, though, I think the trick is to use pedagogical designs that are made to support flexibility, that encourage learners to connect with the spaces live and people they share them with, that support them in understanding the impact of the environments they are in, and, as much as possible, to incorporate conduits that make it likely that participants will share information about their contexts and what they are doing in them, such as through reflective learning diaries, shared videos or audio, or introductory discussions intended to elicit that information. A good trick that I’ve used in the past, for example, is to ask students to send virtual postcards showing where they are and what they have been doing (nowadays a microblog post might serve a similar role). Similarly, it can be useful to start discussions that seek ideas about how to configure time and space for learning, sharing problems and solutions from the students themselves. Modelling behaviours can help: in my own communications, I try to reveal things about where I am and what I have been doing that provide some context and background story, especially when it relates to how I am changing as a result of our shared endeavours. Building social interaction opportunities into every inhabited virtual space would help a lot, making it more likely that students will share more of what they are doing and increasing awareness of both the presence and the non-presence (the difference in context) of others. Learning management systems are almost universally utter rubbish for that, typically relegating interactions to controlled areas of course sites and encouraging instrumental and ephemeral discussions that largely ignore context. We need more, more pervasively, and we need better.

None of this will replicate the rich, shared environments of in-person learning, and that is not the point. This is about acknowledging the differences in online and distance learning and building different orchestrations around them. On the whole, the independence of distance students is an extremely good thing, with great motivational benefits, not to mention convenience, much lower environmental harm, exploitable diversity, and many other valuable features that are hard to reproduce in person. When it works, it works very well. We just need to make it work better for those for whom that is not enough. To do that, we need to understand the whole assembly, not just the pieces we provide.

The physics of social spaces are not like the physics of physical spaces

Posted June 19, 2021, updated onJune 20, 2021 by Jon Dron

Over the last week I peripherally participated in an interesting exchange of views on Twitter between Jesse Stommel and Stephen Downes that raises some fascinating issues about the nature of online social spaces. It started with a plea from Jesse:

“Dear [insert company name], searching every mention of your company and jumping into conversations where you haven’t been tagged or invited is invasive. Stop doing that.”

Stephen took exception to this, pointing out that:

“If I use a company name in a public forum, I expect they will take interest and maybe even reply. It’s a *public* forum. That’s how they work.”

What followed explored some fascinating territory, but the essence of the main arguments are (I skim the nuances), on Jesse’s side, that we have a reasonable expectation of being left alone during a private conversation in any public space and, on Stephen’s side, that there should be no expectation of privacy in a public digital space like Twitter, and that any claims to it tread on extremely dangerous ground. The central question is thus whether there are such things as private conversations on Twitter.

Stephen’s big concern is that, taken to its logical conclusion, laying claim to privacy on Twitter opens the door for outrages like the Proctorio vs Linkletter case, in which Proctorio claimed that “Mr. Linkletter infringed its copyright, circumvented technological protection measures, and breached confidence” by sharing one of its fully public (though not publicized) YouTube videos with students. YouTube quite closely resembles Twitter in its social structure (though little else), so it is a good analogy. Stephen is, I think rightly, concerned at ‘calling out’ individuals or organizations for invading ‘private’ conversations in public spaces because it implies the unilateral imposition of norms, rules of behaviour, and expectations by one individual or group on another, in a space that neither owns.

Jesse’s counter-arguments are interesting, and subtle. He strongly rejects Stephen’s analogy with the Proctorio case because all he is doing is asserting his right to privacy, not abusing his market position or trying to cause harm. It’s just a request to be let alone, calling on what he sees as norms of politeness, not a demand that this should be enshrined in rules or legislation. He observes that, though Twitter is a public space, it has variegation that emerges because of (often tacit, seldom explicit) ways that many (not all) people use it, which in turn is supported by the ways that Twitter’s algorithms push some kinds of tweet more than others. For this particular case in point, he notes that the algorithm tends to broadcast initial tweets more than it does replies, so what follows in a set of replies could be assumed by its participants to be a less public conversation. In fact, as I understand his argument, Jesse thinks of it as a private conversation in a public space, analogous to having a private conversation in a public park where one might be inadvertently overheard, but it would be rude to deliberately listen in or contribute unless invited. If this were a true analogy then I might support it. But, if it is true, then so are quite a few other things, and that’s where it starts to get interesting.

I’ve been a Twitter user for approaching 15 years now and it has never occurred to me till now that any of my conversations might in any way be construed as private. They are sometimes personal, for sure, but definitely not private. Conversations are soft technologies that are flexible, mutable, and situated, and (without further clues like people quietly conversing in a corner) you need to read them in order to know whether you would be intruding on them, which means that they are simply not private. Without further reasons to assume privacy, it is just a conversation in public between two people to which other people are not invited.

So the crux of Jesse’s argument seems to be the notion that a happenstance of Twitter’s current implementation that makes some tweets less likely to be seen than others, combined with a set of norms relating to that, that may or may not be shared by others, allows one to claim that a conversation is not just personal but private.

The physics of online social spaces

Twitter is, as Stephen says and Jesse agrees, for the most part a completely public space (not counting direct messaging or constraints on tweets to only those you follow/are following) but, as the example of the relative prominence given to initial tweets compared with replies to them amply demonstrates, it does have a structure. It is just one that does not obey anything like the same physics as a physical space. You can achieve a measure of privacy in a public physical space because there has to be proximity in space and time in order to communicate at all, and there are limits to human voice projection, ability to hear, and ability to attend to multiple conversations at once. There are also visual clues that people are talking privately. Though there is variegation in structure, none of those limits apply in Twitter or, for that matter, most online social spaces.

Early in the conversation I chipped in to observe that one of the many differences between private conversations in physical space and Twitter exchanges is that tweets are persistent. They are a little like graffiti left in public spaces that continues to communicate long after the initial intent has passed, and may be happened upon at any time in the future in quite different contexts than those imagined by the graffiti artist. Jesse’s response to that was that there’s a difference between graffiti on a public building in five foot high letters and graffiti on a shady tree or in a tunnel. Again, his point is that there are parts of Twitter where there might be a reasonable expectation of relative privacy, where it would be rude to join the conversation. Though I agree that it is often possible to tell from reading a conversation whether you might be welcome or not (and yes, social norms apply to that), my big problem with Jesse’s argument is that proximity in Twitter-space is not just defined by relative position in a dialogue or likelihood of appearance in a Twitter feed, as he seems to imply.

Beyond its support for conversations between individuals, Twitter embodies two distinct but overlapping social forms: the network and the set. @mentions in Twitter combined with its ‘following’ functionality are the main drivers for the network form. If you follow someone or they mention you then your message becomes proximal to them. That’s a big part of Twitter’s physics, and it has no analogue in physical space. Thus, your conversation is very likely to be overheard by others because you are (metaphorically) standing right next to them and chipping your words in five foot letters in stone where they can and will be found, now and in the future. If you wanted to have a private conversation in a park then you wouldn’t stand less than a metre away from someone that you didn’t want to listen in and shout in their face. But that’s not all.

Hashtags and search terms are the main drivers for the set social form, which at least closely competes with if not exceeds the value of social networks in Twitter. When you use a hashtag or even a distinctive word (say, the name of a company or person) then your message becomes proximal to those who follow that hashtag or who have saved a search for that keyword. So you are not just standing right next to everyone in your social network, but to the potentially much larger social set of people who are interested in keywords that you use in your conversation. Again, you might not intend it, you might not even be able to see them, but you are shouting in their faces.

Maybe you do have a right to privacy in any public space, but that right does not overrule simple physics. You have to know the physics of that space in order to know what ‘private’ means within it. And the simple physics of Twitter means that ‘next to’ and ‘within hearing distance’ extends to anyone with an interest in you or what you are saying in the sentences you write. If you want different social physics that support privacy, then you need to take your conversation to a different space, because Twitter doesn’t work that way. You can ask for non-interference in a personal conversation, but not for privacy.

Designing better social physics

As it happens, we grappled a lot with issues of context and privacy exactly like this when we designed the social physics of the Landing. Its social physics are deliberately designed to make precisely those nooks and niches that Jesse wants to find in Twitter. The Landing starts with discretionary access control for every post and every profile field (we chose to build it using the Elgg framework because of its support for this). Like the much missed (and never hit) Google+ it also allows you to create circles, that are not just useful for following but, more significantly, for limiting access to particular individuals. Again, that came for free with Elgg, though we added some enhancements to forefront it, and to make it usable.

It’s not just about the content, though; it’s about presentation of self (we were influenced in this by Goffman’s dramaturgical analysis). We also therefore built a range of context-switching tools – notably tabbed profiles and pinboards (known internally as ‘sets’) – that allow you to present a completely different facade to different circles, groups, and sets of people. This is not just concerned with showing or hiding different fields and content, but with looking completely different and showing completely different stuff to different people. The public facade of my profile is not the same as the one displayed to my friends and, if I wished, I could present different facades to all the different circles or groups of people I follow or belong to. We’ve still not solved the temporal issue – like most social sites, the fundamental unit of communication is still persistent graffiti. In fact, to a large extent we wanted it that way, because it’s a site for collective learning, and so it has to have a collective memory though, like memories in brains, it would be useful to have short-term memories too. However, simply letting posts expire is not the solution, in part due to the many ways that digital content can be copied and archived but, more importantly, because forgetting is and must be an active process that cannot and should not be automated. My earlier CoFIND system did have a way to deal with that (memories had to be actively maintained by active interest and use by members or, though they would never be fully lost, they would be far less likely to be recalled) but we didn’t make much use of that idea on the Landing, save in isolated pockets, because it would have really irritated the many people or groups that engage intermittently (e.g. in iterations of paced courses).

Unfortunately, most of the Landing’s context-switching features are not even slightly intuitive (especially to those already familiar with the cruder social physics of popular social media) so most are very rarely used. Google+, with its massively simplified version of the same idea, probably failed at least in part for this reason. Such complexity can work, with the right membership. Slashdot, for instance, has an extraordinarily rich and ever-evolving social physics, and it has thrived for about 25 years, but the reasons for its success probably lie at least in part in its tagline ‘News for Nerds’. Its members are not phased by complex interfaces, and it is well-enough designed to work reasonably well if you don’t engage with all the features.

Perhaps a bigger issue, though, is that the richer social physics of both Slashdot and the Landing only work if you happen to be a member. For public posts, like this one, the physics are very much like those of Twitter or Facebook.

For now, the best bet is to use different social spaces for different aspects of your life but, thanks largely to Facebook’s single-minded and highly effective undermining of OpenSocial, there’s not a lot of ways to seamlessly move between them right now while retaining a rich and faceted identity. At least there’s still RSS, which is how come you might be reading this on the Landing (where it is originally posted) or at https://jondron.ca/ (which will automagically then push it to Twitter), but it’s not ideal.

It’s very challenging to design a digital space that is both richly supportive of human social needs and easy to use. The Landing is definitely not the solution, but the underlying idea – that people are richly faceted social beings who interact and present themselves differently to different people at different times – still makes sense to me. As the conversation between Jesse and Stephen shows, there is a need for support for that more than ever.