‘What separates humans from machines is the fact that
we are creative’
Data science has been revolutionising the insights
industry. If we were to believe the media, artificial intelligence (AI) and
machine learning (ML) applications are already fully integrated in countless
industries. But what is it really? Can technology take that next giant leap and
actually drive qualitative research? Will it ever be able to predict human
behaviour and account for biases? In reality, this technology has a long way to
go. We put these intriguing questions to a panel of experts. Spoiler alert:
humans will still be needed, as AI is still rather restricted to humble text
analysis.
Let us first introduce our four experts by asking them
for their own definitions of AI.
Definitions
Bart Langton, research director at Ipsos New Zealand, talked about big data, AI, machine learning and behavioural science at ESOMAR Asia Pacific 2017. His definition of AI is “any kind of automated task that looks at patterns within data. This data can be unstructured and more qualitative in nature, or more structured, like big data.”
Jonathan Williams, founder of the AI driven platform Discover.ai, believes that right now AI is less about machine intelligence and more about machine learning (teaching machines based on example, not programming based on rules). “Most of us are combining machine learning with programmatic technology to create smarter tools that enable agile insight. In short, the definition for us as an insights industry should be something like: AI = smarter machines, smarter insights.”
Kyle Findlay leads the Kantar Innovation Global Data Science Team based in Cape Town and gave a presentation at ESOMAR Big Data World 2017 in New York, where he won ‘best paper’ for his insight in big data, artificial intelligence, data science and deep learning. He thinks the term AI is very amorphous. “It basically means ‘machine doing smart things’, and what is considered smart is in the eye of the beholder and also evolves constantly. At this stage, I’d consider something that uses machine learning to make a prediction as AI, especially if it continues to learn and evolve from its experience.”
Jonathan Mall is CEO and founder of Neuro Flash, which uses patterns and huge banks of words to predict human behaviour and account for biases. He says “AI is the ability of a machine to exhibit ‘intelligent’ human behaviour. For example, the Neuro Flash AI accurately predicts human associations to imitate how consumers perceive brands and content.”
Accelerating human expertise
Challenging as it is to delineate what AI and data
science comprise, we encounter even bigger hurdles when connecting these
technologies to market research, and the qualitative variety in particular.
According to Williams, much of the advances in AI in
the insights industry to date has been about automation and using smarter
machines to take people out of the loop. But, “there is another, more important
role for AI: accelerating human expertise. It’s about recognising the best of
what makes us human and developing smarter machines that work with us to
accelerate and maximise those strengths.”
Findlay believes that as AI improves, machines are
able to do more and more things that used to be the preserve of humans, and do
them at scale. “Historically, qualitative has been the preserve of the more
human side of research – the side that recognises empathy, emotion and the
complexities of the human experience, which it tries to distil down into
relevant insights. Conversely, quantitative research was the preserve of cold,
hard structured data, which has always had its limits when it comes to human
understanding. The trade-off used to be: a loss of the complexity in favour of
scalable, structured data versus less scalable but more complex insight through
qualitative. AI bridges this gap. More and more, we are able to account for the
complex vagaries of the human condition at scale, essentially empowering
quantitative qual research.”
Blurred lines
AI as merely an upgraded text analysis tool is selling
the technology short, most believe. Williams explains where this notion stems
from. “Text analysis is where there is a lot of action because there is a lot
of unstructured data out there in text form. But in fact, to the machine
learning algorithms, text is really no different to any other sources of data, like
analysing a picture or facial recognition.
Indeed, Mall can see AI advancing in many areas. “Using
competing network architectures or deep learning approaches, we are starting to
solve very complex problems. Even long-term strategic planning, short- and
long-term memory and curiosity are already effectively modelled. And we analyse
thousands of consumer verbatims to quickly find content that contains relevant
associations.”
For truly creative AI work, Langton concedes that the
more complex machine learning is developing, and that faster computers and
cloud computing will accelerate things, especially with the likes of Apple and
Google moving in. “We’re slowly but surely competing with them. They use all
their information and data to predict human behaviour. And that’s what we do as
well. The lines between what tech companies, consultancies and traditional
research companies do, are blurring.”
Paradigm
No matter the developments, Williams is not worried
about the future of human researchers. “Qualitative research is about people,
their motivations and behaviours, and the brand solutions we are trying to help
build are creative. Experts in the qualitative field will naturally be resistant
to the idea that machines can do what they can, and in this I think they are
absolutely right.” Give the same insight experts the technology that accelerates
their own expertise, and it frees them to explore more sources of insight more
quickly to get to deeper insights, and Williams thinks AI technologies will
flourish. “We’ve seen it happen already and the potential for the future is
huge. So, expect more acceleration.”
“To the extent that we still engage in interviews,”
explains Findlay “they will probably end up feeling more like a one-on-one
conversation between a therapist and a patient than the rote, staccato
interviews of today. These conversations might not even be with real
interviewers but rather chatbots, robots and assistants. Beyond this,
personality, motivations, etc., will likely be derived indirectly from
observational signals and used to customise experiences, as we see with
programmatic advertising and personalised offerings. Such understanding might
be put into a qual-inspired framework to make it understandable and impactful.”
Democratizing
What will be automated first, predicts Mall, are
annoying tasks, like response coding. Here, AI agencies will be able to
outcompete on price and speed. But with cultural and behavioural tracking
becoming a common thing in online and offline content, he also believes that
the future holds many exciting ways of integrating AI in the research process.
Langton predicts huge applications, especially in handling
more dynamic conversations humans have with each other, facial recognition, and
decoding volumes of ethnographic video and derive meaning from those. “It might
be able to uncover more meaning or ‘truth’ behind what people actually mean,
based on their facial expressions and tonality. So that’s probably where
qualitative will go with AI in the future. It’s just that I’m not sure how far
away that future is and how fast things will move.”
Williams sees a lot of apprehension, fear and
scepticism around the change that AI can bring. But he says: “If we all work
together for a future that embraces acceleration, as well as automation, it
will be a positive benefit to the whole industry. It’s a really exciting time
to be in the industry.” Mall believes it’s time to stop seeing qualitative and
quantitative as opposites. “In the end, both rely on data to draw conclusions
and advance our understanding. AI may be one way of enabling people to see that
combining data in meaningful ways will benefit everyone.”
This article is an edited excerpt of the original “What is data science after all, and how human is it?”, published in the 2019 Global Market Research report. Read the full content by accessing the report, here.