Saturday, November 22, 2008

Honest Signals: How They Shape Our World

Author: Alex Pentland

A 'nervous system for humanity'? John Gilbey finds a sting in the tale

A slender hardback book is sitting innocuously in a pool of sunlight on my desk. With fewer than 200 pages, of which almost half are taken up by appendices, you might not expect it to carry the seeds of social revolution - but I strongly suspect that it does.

In Honest Signals: How They Shape Our World, Sandy Pentland and his research team from the Massachusetts Institute of Technology Media Lab examine the different ways we communicate within groups. This may not seem, at first glance, to be anything startling or novel. After all, anyone who has been subjected to a management training course in the past few decades will have been instructed in the importance of body language and non-verbal communication at some level, however basic.

What is innovative about this tranche of research is the development of a set of technology-based tools to enable the automated capture of behavioural information in an intensive and robust way - providing a demonstrable degree of numerical rigour in support of their conclusions and dramatically extending what is possible in experimental terms.

The concept of honest signals is explained by Pentland as those elements of communication and display that are processed by us unconsciously - or are effectively uncontrollable, or are difficult to fake - so that they can provide an intrinsically valid stream of data with which people guide conversations, meetings and decisions. He isolates four examples of honest signalling for closer study: influence, measured as the extent to which someone modifies the pattern of speaking of another person to match their own; mimicry, or the way we copy the behaviours of another through the course of a conversation through smiles, comments and nods; activity - how interested and excited you are is apparently reflected measurably in your level of activity; and consistency - Pentland suggests that consistent levels of emphasis and timing in speech indicate mental focus in the speaker, with any inconsistency leaving us open to influence from others.

Critically, these honest signals can be read clearly - by people or gadgets - in a variety of environments where the finer-grained information of language and expression might be lost: think of discos, bars and crowded streets. Pentland views this as a connection to our remote past - when accurate communication around the campfire or in the depths of a forest was a key survival element.

Forming the core of the book is a discussion of the work carried out by Pentland's group over the past five years, showing the evolution of the research programme. Inventive data capture and analysis, using novel in-house developments and the manipulation of existing technologies, has enabled the mass observation of subjects, generating thousands of hours (330,000 hours is quoted) of data. In the narrative, the various data collection devices are referred to collectively as "sociometers".

The current version of the sociometer is a sophisticated but unobtrusive device that can be worn like an ID badge. It can tell how much time you spend talking face to face with a named person, carry out speech analysis to measure social signals and social context, recognise common activities by measuring body movement, determine where you are in the building, and talk to mobile phones and computer networks to exchange data and measure exactly where you are in relation to other people. Try doing that lot for a roomful of people using a clipboard and a stopwatch.

One slight disappointment is that the tools developed by the research team are not described in great technological detail in the book, so geeks like me have to slide over to the website for more information.

Most of the team's research papers are available in PDF format - but be warned, the website (http://hd.media.mit.edu) is one of the most garishly presented I have ever seen.

Pentland builds the story upwards from a fairly traditional discussion of group roles, and the interactions within teams, through the challenge of "reading" poker players - real and metaphorical - to the questions around how the power of the group can be most effectively harnessed. Strangely, the linear discussion in the main text becomes almost secondary to the material carried in the compendious appendices, which relate how each component of the argument was tested.

I found myself repeatedly flipping back and forth between chapter and support material - but with deepening interest rather than frustration. The appendices are, in many cases, mini-research papers with a hypothesis/method, results and discussion format forming a clever bridge between guidance for lay readers and the more formal expectations of a scientific readership.

The analysis and reporting of the sociometer data makes a persuasive case for the use of this technology in the understanding of social networks, but there is a lot more at stake here than that.

The technology offers, in Pentland's words, a chance to "enable a magnification of our social sense", to step outside our own behaviour and observe it dispassionately as a set of statistics. On a substantially larger scale, it potentially provides a set of techniques on which to base the development of a "social physics" - for which the author provides a convincing proof-of-concept analysis. Readers of Isaac Asimov's "Foundation" stories will be forgiven if they feel a hint of Hari Seldon's discipline of "psycho-history" at this point.

Throughout the book, the methodologies and the academic treatment of the subject are impressive, but there is - for me - a significant sting in the tail (or tale, for that matter). In developing the data-capture technology, the team has sensibly used standard components wherever possible. One of the team's main platforms has been a series of increasingly powerful mobile phones, with Linux operating systems and Bluetooth communications, for which it developed some interesting applications.

The software described in the text and in the supporting research papers is intriguing - Meeting Mediator, for example, uses the sociometer data to provide immediate feedback to enhance group collaboration, feedback that is presented on the screen of each phone, thus "encouraging" participation. My particular favourite, however, is the Jerk-O-Meter - yes, really. This application provides appropriate feedback via phone screen to someone whose interest in a phone call is perceived, by the system, to be waning.

Clever stuff, but this is surely appropriate only in the context of a development tool and an academic exercise. The problem I foresee is that it will be only a matter of time before big organisations decide to include such features by default in the phones used by their staff to provide them with daily (hourly?) reports on how much each person is perceived to be contributing to discussions, and how attentive they are to their phone conversations. Heck, you can see where I am going with this.

In the epilogue to his book, Pentland presents the case for using these developments in the creation of a "nervous system" for humanity, evolving into a structure that enables us "to engineer our societies and entire culture".

The author is obviously aware of the risks this could pose to privacy and individual liberty - and this book provides a timely wake-up call so that we can decide where the boundary lies between "possible" and "desirable".

For once - just once - please let us have that debate in public before the whole issue is dismissed as "inevitable".

The author

Alex (Sandy) Pentland is the Toshiba professor of media, arts and sciences at the Massachusetts Institute of Technology. Specialising in human-centred technology, he is also passionate about creating ventures to take the technology into the "real" world.

To that end, he has founded or co-founded half a dozen institutes and set up ten spin-off companies. His achievements have been recognised not only by his peers - who have cited him so often that he can now claim to be one of the top-cited computer scientists in the world - but also by the mainstream media. In 1997, Newsweek named him one of the 100 Americans likely to shape this century. During his career, he has taught at the University of Rochester and Stanford University and has worked for two non-profit organisations.