Doing some research on this soon and it’s pretty new to me. Here’s an interesting article that explains the basics for anyone interested: Tech Crunch: machine learning
I’m pretty confident it will be a long while, if ever, before machine learning usurps the proper research consultant as the deliverer of insights. But it’s worth thinking about what kinds of qual data gathering may in future be enhanced by forms of machine learning – web-scraping springs to mind – to give us more tools, or indeed to perhaps reduce our role in certain types of data gathering. However, qual is more protected than most from being simply dumped out of a job by this technology:
- our role has long been partly as a “reality check”, a complement to the “desk” data that shows clients whether what the data seems to be suggesting about people’s thoughts and behaviours is really the case in real life. In a machine learning age, that is no different.
- much of our data generation is from human interaction, usually face to face – people often need social interaction, prompting through ‘natural’ conversation, to open up and share what’s interesting. We’re the masters of that.
- qual is participant-led – that is, we’re not throwing batteries of questions at people, we let them lead while we gently steer the topic flow. We respond to particular comments and probe on what they mean, both individually and in relation to what other people are saying. This isn’t something machine learning could easily be applied to, even if we can imagine a robot moderating a group (and that’s some leap, for now at least).
- In the analysis, we listen not only to the literal meaning of words, but also need to assess their meaning in several contexts (e.g. as part of a story, in a physical environment, within the wider language, within the wider culture, their meaning in brand terms, in business terms, etc) – not to mention the participant’s tone, body language, mood and so on. We jump between these islands and these layers of meaning to reach holistic insights that draw it all together and make sense of it, to create advice for our clients. Again, for machines it’s the stuff of artificial intelligence, way beyond where machine learning is now or is likely to be any time soon.
I don’t want to sound complacent – who knows where we will be in 20 years – but it does strike me that qualitative research data gathering, analysis and insight, being about as far away from mechanistic thinking as it is possible to be, is one of the least machine-replaceable pursuits there is.
We in qual are a bit different and a bit removed from the usual worlds of our clients in business, government or NGOs. We sit partly in their world but we move to the rhythms of the wider public more than the rhythms of our clients. It’s been thrown back at us as a weakness. But this, I passionately believe, is our very strength – the source of the real value to our clients. Being a human-shaped practice could be what keeps us relevant and needed, not only in the face of the rise of technology but because of it.
Qual is a messenger from the real, messy human world, into the sometimes sterile environments decision-makers find themselves inhabiting. The messy and the human have plenty of life left in them yet.