Big data (*) – power of correlation of patterns without necessary causation, understanding or explanation – makes sense when the data largely reflects human psychology and behaviour. Because the human explanations would involve rationalisations and game-theoretic responses, not objective scientific causation. The what may be useful even before the why is considered. The what may provide pragmatic value, even near-real-time usefulness, whereas the why may ultimately provide new “knowledge”. These are not mutually exclusive.
The word model is overloaded here – all models are wrong, but some are more useful – but “useful for what?” matters. Important to recognise the different kinds of “model” here and how they’re used – statistical patterns and correlations of what happens, to assign odds to predictions (to markets, say) as distinct from models of mechanisms and processes, that represent how the world works (the economy, say) – especially in the latter case where we are trying to decide inputs, apply agency, to achieve valued outcomes.
Newton:
“I can model the stars in the cosmos,
but not the madness of men.”
Humans will respond to the outcomes of such models – it’s a never ending arms-race – where values are at stake. The usual adage about management by measurement distorting the human process.What can be measured crowds-out other values that matter more to humanity.
The sceptics here are Tiffany and Lisa (and me). Lisa even makes the slip of referring to the geeks in maths and economics as “men”. Many a true word.
(Fair bit of discussion about brains, minds and consciousness too … even shared consciousness beyond any given brain … a longer story.)
[(*) And when I use “big-data” here, I’m simply implying a sample size so huge, that it is likely to be statistically significant for any pattern discovered within it, as opposed to a sample gathered with any particular investigation in mind. Like “cloud” it’s rapidly becoming just the latest jargon for the web of information on the internet of technology.]