It will be a long time before machines truly understand the complexity of the world and how we perceive it – which is why media research still needs humans, writes Craft’s Konrad Collao
It’s been an interesting time in the world of media research (and research more broadly). Mediatel’s Media Research Awards, Future of Media Research conference and Connected Consumer Awards have passed, with the Connected Consumer Conference imminent.
We’ve also seen the publication of the British Polling Council’s inquiry into the polls’ failure to predict the 2015 General Election.
Following these events closely it has become increasingly clear that we are living through a critical period in media and media research. We should all be asking ourselves: how can we know? It seems like the kind of esoteric question asked by earnest undergrads in halls around the land, but it’s fundamental to media.
Clients buy and media agencies sell according to what they commonly agree they ‘know’ about audiences and users, consumers and respondents, or whichever other label we choose to describe the people we are interested in reaching.
In this context research exists purely to further that knowledge, be that to prove sales cases, to truly explore new behaviours or to simply give the industry a picture of what the hell is going on ‘out there’.
I was particularly struck by a debate about the effects big data will have on media research. We all know that exponentially increasingly amounts of data are being collected. The received wisdom seems to be that this is an unequivocal boon.
The fashion has been to signal “the end of the polymath” as agencies specialise, with the most anticipated specialisation being in data analytics. A particularly striking comment from an industry grandee was that some of the more innovative observational methods – full disclosure, that we at Craft specialise in – while interesting, will become “peripheral”.
Looked at one way, we can see merit in this argument. Ultimately media is all about selling. To set a price you need to measure, to prove and convince: reach, frequency, ROI, ‘engagement’ (whatever that is – can anyone define it?).
With more accurate ways of measuring, with more measurement data than we know what to do with, with people better versed in mining it, the major existential question we at Craft ask ourselves is: what is the role for observation and intuition in an age of supposedly incontrovertible ‘proof’?
Anyone who has faced a quantitative dataset, especially derived from an ad hoc study, knows that one of the first questions to ask yourself is: where do I start? It’s not always clear what you are looking for, where you might find it or once you have found it, what it means.
So despite quantitative research’s reputation for being ‘objective’, ‘rational’ and other adjectives that convey the impression that numbers cannot be argued with because they take ‘subjectivity’ out of the equation, we believe that humanity can never be eliminated. Because the human can’t be eliminated, nor can intuition.
So, is that a bad thing? Undoubtedly, not recognising the effect of intuition and biases (we all have them) on research design and analysis is problematic. But if one recognises that the very act of researching – of setting hypotheses, designing an experiment and analysing its results – makes the researcher present, then we can mitigate those effects.
Even that phrasing suggests the presence of intuition and humanity is problematic, that what we ideally need is a machine. The problem is that machines aren’t very good at understanding humans for the foreseeable future at least.
Yes, they can beat us at Go. Yes, they can drive a car (sort of). But it will be a long time before machines truly understand the complexity of the world and how humans perceive it, which is what media ultimately plays on. ...
… read on at mediatel.co.uk