How bad is my filter bubble problem? Please help me find out

January 6, 2017

Links I Liked

January 6, 2017

Book Review: Social Physics : How Social Networks can make us Smarter

January 6, 2017
empty image
empty image

My Christmas reading included a book called Social Physics – yep, a party animal (my others were Lord of the Flies and Knausgard Vol 3, both wonderful). Here’s the review:

Airport bookstores are bewildering places – shelf after shelf of management gurus offering distilledsocial-physics-cover lessons on leadership, change and everything else. How to distinguish snake oil from substance? My Christmas reading, based on a recommendation from someone attending a book launch in the US last month (thanks whoever you were – all a bit of a blur now) was Alex Pentland’s ‘Social Physics: How Social Networks can make us Smarter’. I must confess, as a lapsed physicist, the title swung it for me, but I learned a lot from this book. At least I think I did – let’s see if I am still using the ideas in a few months’ time.

Social Physics is not a new idea. Auguste Comte, the founder of modern sociology, coined the phrase back in the 19th century. Comte and his crew aspired to explain social reality by developing a set of universal laws—the sociological equivalent of physicists’ quest to create a theory of everything. As with economics, that kind of physics envy has proved largely delusional. Now though, Pentland argues that the arrival of Big Data means we can aspire to a ‘thermodynamics of society’, where behaviour is governed by discernible mathematical laws. It does not deny free will – Pentland does not claim to be able to predict individual behaviour, but finds a high degree of certainty in mass behaviours, which appear to follow particular patterns (like atoms in a gas).

Big Data allows us to move from describing society in terms of stocks (equilibria, population, education and health status) to real time flows (of information, ideas, contacts) and this transforms are ability both to understand and accelerate human progress.

The two most important concepts in SP are

  • Idea flow within social networks, and how it can be separated into exploration (finding new ideas/strategies) and engagement (getting everyone to coordinate their behaviour)
  • Social learning, which is how new ideas become habits, and how learning can be accelerated and shaped by social pressure
Lots of data

Lots of data

Compared to other gurus with similar messages about the need to move beyond assumptions about ‘rational, utility-maximising individuals’, Social Physics has more emphasis on social interaction and collective rationality than Kahneman or Nudge.

Pentland sees the flow of interaction and ideas in social physics as ‘a language that is better than the old vocabulary of markets and classes, capital and production’. The tone is ambitious: ‘we can begin to build a society that is better at avoiding market crashes, ethnic and religious violence, political stalemates, widespread corruption, and dangerous concentrations of power.’ Sign me up.

The later sections of the book on how to redesign cities (we need to be more like Zurich, apparently) and rethink politics and governance for an era of Big Data were a bit underwhelming – what got to me were the earlier sections on individuals and organizations. And I found some of it quite uncomfortable.

According to Pentland, creativity is born of two processes: the first is exploration, where people move out of their comfort zones and actively seek people with different views and ideas. He also stresses the importance of curmudgeons and contrarians who argue against the conventional wisdom of the day. Great, that’s what I spend a lot of my time doing – reading books like this for example, or talking to academics and businesspeople as well as activists, or pouring scorn on the latest fads and fashions when they annoy me.

The second process is engagement, which proved a lot less reassuring when I think about how I behave in meetings. ‘Groups have a collective intelligence that is mostly independent of the intelligence of individual participants’. The most creative and effective processes of engagement are those that show ‘equality of conversational turn taking’ – discussions in which no-one dominates, lots of short idea-and-response type exchanges. The key finding is that the dynamics of the exchange are more important than the knowledge and experience of the individual participants. The exception is ‘social intelligence’ – the ability to read other’s social signals, in which (surprise surprise) women have a distinct advantage.

So what are the practical implications? Well, one of Pentland’s numerous spin-off companies can

Which one's me?

Which one’s me?

rig you up with an app to monitor meeting dynamics: a visual display on your phone shows who is hogging the airtime – particularly useful when some participants are not in the room (eg conference calls) or speaking a second language). In another case they persuaded a company to make its lunch tables longer, so people chatted beyond their core peer group. Coffee pots and water coolers (hell, maybe even Open Plan) turn out to be essential. More generally, you can think about the balance of exploration and engagement (one of my many unheeded suggestions in Oxfam is that everyone should have a read-out of the percentage of their monthly emails that comes from non-Oxfam sources). That balance needs to be at both an individual level (everyone probably needs to get out more and ‘oscillate’ between exploration and engagement, even if it’s just having a coffee with people from elsewhere in the organization) and collective level (how is our balance of explorers and engagers?) Some activist organizations have a monthly target for one-to-ones with grassroots leaders – maybe aid agencies should do something similar?

These are not terribly original insights, but Pentland’s argument is that Big Data allows you to precisely measure and optimise an organization’s performance, ‘tuning’ its combination of exploration and engagement. As proof of the pudding, his spin-offs have done this for a number of companies and others with (he says) good results. I’m torn by this – sceptical that you can apply such quantitative methods to something as human as an organization, but also attracted by the focus on systems and interactions, and the lure of promised precision.

Any other caveats or qualms? A couple. Even by my narcissistic standards, the book is remorselessly first person: ‘I’ is probably its commonest word – Pentland, who is a prof at MIT, does not suffer from false modesty and locates himself at the heart of pretty much every important idea or development in the field. Then there’s ‘we’, as in ‘we can build better social systems’. Who, pray, is ‘we’? Who gets to decide who can join? What motivates the ‘we’? – the pure altruism of the disinterested datacrat, or something much more complicated? He seems remarkably un-self-aware in all this.

Which, you won’t be surprised to hear, makes me wonder about power and politics. Sure, Pentland has some quintessentially American ideas of curbing state excess through a ‘New Deal on Data’ that combines safeguards with the benefits of openness, but otherwise, this is a datacrat’s dream, with philosopher kings wielding algorithms and apps for the benefit of the common weal. Published in 2014, I wonder whether he would change much in the era of post-truth politics?

And here’s Pentland introducing his ideas (50m, he starts at 2.12)

3 comments

  1. Hi Duncan

    Does this guy acknowledge, even in passing, all the prior work done by others? The field of social network analysis is absolutely huge, spanning many disciplines, and has had annual international conferences for many years (http://www.insna.org/#) Even the distinction between exploration and engagement sounds like his twist on the widely used distinction made by March years ago, between exploration and exploitation (http://www.analytictech.com/mb874/papers/march.pdf)

  2. Hi Duncan

    Thanks for your great analysis of something that constitutes the sea I’m swimming in! I work with data scientists and am excited by the journey of figuring out what the heck is useful. My PhD research in Informatics was is ICT4D and focussed on individual entrepreneurs using the internet to make a living in rural areas. I investigated their social capital to make sense of social networks in action and how it helps and constrains. I tried to describe the context of bottom-up development and top-down development using Max-Neeff’s work. I ended up with a rich description of a system of relationships based on interviews with 14 people.

    As Rick Davies pointed out a whole host of sociologists and development researchers have been working in this kind of social systems work. I am trying to make sense of the balance between people created networks and the meaning it has for them and the data scientist view of the myriad of relationships (many of them teneous) that are now possible via social media and the emergent patterns of interaction. I think the structural aspects of the networks is the easy bit. I’m curious about the cultural and other sense-making aspects and wonder who has combined the textual analysis of the conversations with interpretations by members of the social networks? I know of Dave Snowden’s Cynefin work and software, but as you say, companies are now appearing that promise what seems to be very neat packaged solutions that makes me uneasy about the major hidden assumptions.

    Many thanks for the post!

    Regards

    Mario

  3. Hi Duncan, from a bit of experience participating in ‘big data’ discussions with Sandy and colleagues (albeit some time ago), I share the mix of excitement and reservations that you voice towards the end. Datacratic views often don’t do much acknowledging of the human and the political – implicit judgments, the governance side of things, differences in capacities between parts of the world. The thinking may be changing though – some in the big data field are very aware of these issues.

Leave a comment

Translate »