Open Intelligence
On: 7th March 2012
Q: Can a general theory of intelligence help us to better understand what we're up to with this whole open data business?
A: Yes. Yes it can.
A: Yes. Yes it can.
Some rights reserved by cuellar
Last Thursday's Opening Doors event brought the conversation about the voluntary sector's relationship with open data to a wider audience, generating (amongst other things) a set of pragmatic requests, a call for greater leadership and a thoughtful summary of an 'Open Data Day'. My own contribution is to suggest that David Krakuer's analysis of intelligence, set out in this series of three brilliant lectures, can offer us some important insights into exactly what's on the end of our open data fork (and why we want it to look like a roman cauliflower).
Krakuer suggests that intelligence emerges wherever four processes are taking place: representation, inference, strategy and competition.
Representation
A field mouse sees a tasty seed, and takes in its surroundings...
Inference
...then predicts the outcome of various courses of action based on experience and instinct (which is derived from the cumulative experience of generations of field mice)...
Strategy
...then chooses a particular course of action...
Competition
...then either gets the seed, loses out to a cunning shrew
or gets eaten by an Owl.
Krakuer goes on to present the evidence for what he calls Cognitive Ubiquity. This is the idea that intelligence in this sense is identifiable at multiple levels, from the actions of a human being down to a single neuron within that human's brain: 'individual' intelligent organisims are in fact made up of similarly intelligent elements and systems in which those four components are observably in play. So our digestive system, our stomach, the cells in the wall of our stomach all exhibit these same basic characteristics of intelligence: they form a representation of their environment, 'choose' an action from a range of options, and exist in their current form only because they performed their function more effectively than the competition.
It is tempting to claim that we can extend this spectrum of intelligence beyond the individual human: families, organisations, societies, nations can surely in some sense be described as behaving in a structurally similar way. However, I believe that this train of thought is primarily useful in that it pinpoints some ways in which we, as collections of interconnected individuals, fail to achieve the eloquent, evolved intelligence exhibited by biological systems.
I'd also argue that one of the keys to this failure is our troubled relationship with data: a relationship that might be improved by making more data more open, but only if we attend to some underlying issues, at least some of which can be brought into focus using Krakuer's analysis.
How do the staff, volunteers and trustees of a voluntary organisation 'represent' their work to themselves, to their funders, to the people they support? Data, open or otherwise, will play a part, but this is always coloured by more diffuse information, primarily in the form of personal experience. Being personal, the impact of such experience on what an individual actually 'sees' is likely to be significant: yet, as groups, we struggle to accommodate the "anecdote", often feeling that it is some kind of quaint bartering token up against the 'hard currency' of data (which is numbers, spreadsheets, all that kind of stuff, isn't it...?).
Sadly, if we get off to a bad start, the rest of the process is compromised: a partial, incoherent representation will impede an organisation's ability to draw useful (i.e. predictive, accurate) inferences. More troubling still is the likelihood that an organisation's representation of itself is further distorted by a predetermined strategy: 'policy-based evidence-making' to borrow a phrase from the more cynical (or realistic) corners of the civil service.
The operation of competition is where things get really messy: individuals and groups compete for resources, connections and support, arguing their case through constructing a representation of themselves which they believe to be what their audience wishes to see. Data is trimmed, polished, flattered with language and, where deemed advantageous, seasoned with carefully selected anecdotes: which in turn sets off another compromised, partial process within the target organisation (made up of individuals who all bring their own more or less arbitrary baggage to the party).
All of which adds up to a collective delusion of intelligence: we obscure crucial subtleties through compromise, hacking away at the mess to 'reveal' something solid, objective, unarguable. Success is defined in self-serving terms by those with the most power, confidence or ambition, fatally distorting the inexorably improving process of genuine competition.
I'm being deliberately catastrophic. We are still limping forward as a species, but I believe this to be primarily a testament to the mitigating influence of our finer instincts over the sophistry of our public discourse.
And I believe that the momentum building behind open data has its roots in those finer instincts: an intuitive sense that honest, exhaustive representation is the crucial starting point for a journey towards a truly intelligent society. We are at a soberingly early stage, and the challenges we face are numerous and profound: most pressingly, we must both expand our definition of data and find ever more eloquent, coherent ways of collating and presenting this data to ourselves.
Opening data is merely a first step in highlighting the weaknesses and gaps in our understanding of what we're really looking at. The next steps must now be to address those weaknesses, and work towards a consistent, cumulative representation which builds upwards through the levels in any given hierarchy. A truly intelligent society will display the fractal intelligence of biological systems, in which the processes of representation, inference, strategy and competition can be seen acting at all of these levels, interlocking and combining to form an integral whole.
Split apart a roman cauliflower and you get lots of little roman cauliflowers: an openly intelligent society will be similarly divisible.
__________________________________________________________________________
If you think I'm talking a load of brassicas, let me know - leave a comment, send me an email or find me on twitter @ejanderton.
Thanking you.
Comments
Dynamic Systems
Hi Ed, I think you're right about these issues, but just a couple of points (sorry...longer than intended!).
I think the thing I'd say on this is that what we don't want to do is to assume there's a "true" big picture out there which we need to gather - the field mouse's picture is activity oriented; the mouse is hungry, its 'representation' is towards that end. What we need to conceptualise is the things we're aiming at, and the factors which might be important.
I'm also a bit uncomfortable with representation analogies although it's probably less relevant for this. (These psychs keep recommending the reviewed book, and their blog's good too on this). Anyway, I think Krakuer is using representation as an analogy in the same sort of way as people like Dawkins talk about 'selfish' genes or 'doves' v. 'hawks' - i.e. it's not about individuals building an internal image of an external world, it's a sort of analogy for describing behaviour-in-activity/a particular ecology. The bit that probably does matter here is that what probably happens is more dynamic, and automated.
That is, some systems have particular affordances, how individuals act within them isn't about them building a representation of the system (except in the sense I think Krakuer analogises - that our genetic progress, perhaps our historic sense of design, etc. may indirectly 'model'). What it's about is individuals making use of the affordances of particular systems, and the sorts of outcomes we're hoping to get from those behaviours I guess. So good website design is a nice one, rather than gathering all the data we can, it's more interesting to look at what sorts of change impact on useability (by some measure or other) - and that's what individuals do too, they're not building a representation of the site, they don't need to, it's out there in the world, they're engaging with activity-potentials afforded by the environment, and what they do has a dynamic effect on their subsequent behaviour often in bi-directional ways.
Patterns of salience
Simon, thanks for such a considered and insightful response.
You're absolutely right about the non-existence of the "true" big picture. I largely buy Kant's take on the phenomena-noumena distinction, which places the totality of what is permanently out of the grasp of our senses and our reason.
Which, as you point out, isn't too much of a problem most of the time, as our sense and reason are almost always directed at some more-or-less well-defined goal, following a pattern of salience in our environment.
So a user of a website will choose their path according to their initial intention and the series of options given to them at each stage. Their actions are the result of an interplay, a variety of performance.
The same is true of the designer reflecting upon a website's useability, as they define the term's meaning and choose salient measurables which capture that meaning.
This is the 'some measure or other' bit, which I think gets interesting. Is a term like useability most likely to be defined by looking at what measures we already have - or could easily automate - rather than more costly, messy qualitative stuff? Clearly both happen, but in terms of the volume of users considered there'll never be any competition.
In the short term I'd argue redressing that imbalance is impractical, but I think we can and should be more creative in the way we combine the two to encapsulate meaning.
Yup!
Absolutely, and there has been some progress on those combinations including in Learning Analytics, and with some companies looking at more sophisticated ways to parse qualitative data (I know Xerox are doing this, for example).
Intelligence and politics...
Hey Ed,
This is a really interesting exploration. I found I wasn't entirely comfortable with the open data - intelligence framing. I think it brings something, but I found myself reacting against it. I'll try and explain why below...
Simon's point that there might not be a 'big picture' we can uncover is important here. The point can be read two ways: (1) as I think Ed has taken it, that we can't **see** the big picture; but the other (2) would be that **there isn't** one big picture at all. Particularly not when we come to many of the social issues third sector organisations are dealing with.
The big picture is contested: with the resources we have, is the world we want to create one in which we put our resources here, or there? In this sense, I think open data can be about more informed decision making, but I'd be cautious about framing it as intelligence.
The other grounds of that caution is that we can risk making out open data as an a-political thing if we focus on it as driving rationalist decision making. Digital data is always a reduction of the world, and that reduction is usually shaped by particular world views of the data collectors / political powers (e.g. the choice of Ethnic Monitoring categories an NGO might use / the boundaries placed by local authorities). Once we're a few steps from raw data, different datasets don't naturally fit together perfectly, but need human synthesis - informing, but not constituting intelligence.
This said, I think if we take theories of collective intelligence put forward in works like Wisdom of the Crowds (which outlines the very specific set of circumstances where crowd-sourced decision making does work), then open information can potentially support forms of distributed/collective intelligence.
Absolutely agree...
...about open data and intelligence not being comfortable bedfellows - I think what I was trying to convey was how far away human political/social analysis - based on data to a greater or lesser extent - is from the evolved intelligence in which Krakauer is interested.
In terms of 'big pictures', I was being a bit metaphysical - the knowable/observable versus an underlying reality which we can never fully comprehend. I agree that in a more practical sense 'big pictures' are based on partial, political assumptions - and we should be making those assumptions as visible and open to challenge as possible.
The roman cauliflower image was an attempt to communicate the possibility of having pervasive, consistent transparency throughout a hierarchy, with both data and the analytical/decision-making processes in which they've been used being available to anyone who is interested. So the Sec of State for Work and Pensions could dip into the work programme materials and find a video diary uploaded by a participant, while the participant (or, more likely, the subcontracted vol organisation and/or employer they're working with) could access the meta-analysis done within DWP.
So I'm not necessarily talking about crowd-sourced decision-making: it's more that creating 'open' (relevant, consistent, accessible...) data sources and analysis might tend towards more grown-up, nuanced decisions being taken at different levels. Well, a bit of hopeless optimism never did anyone any harm! :-)
relevant blog
Just had a look at this post (from a Guardian blog) on Big Data and I suppose just how big it is, or how representative, etc. Quite relevant http://www.zerogeography.net/2012/03/big-data-and-end-of-theory.html
Add new comment