What researchers are asking about AI now: Human Insights Conference

AI came up throughout The Research Society's Human Insights Conference in Sydney, with much of the discussion focused on how researchers are actually using it.

Adam Spencer opened the conference with "The GPT Revolution," including what happens to our own skills as AI takes on more tasks. Andrea Clarke's "Building our Adaptive Intelligence" focused on curiosity and adaptability.

Using AI is quickly becoming part of the job. So is knowing what to ask, spotting when something doesn't look right and deciding what deserves further investigation.

Using more of the research we already have

Amelia McVeigh of Lion spoke about something I hear often from client-side research teams.

In "Beyond the Prompt: AI as a Research Collaborator," she described the change in perspective that came with moving from agency to client side. Instead of concentrating on one project at a time, she was now looking across years of research covering different brands, categories and business questions.

Years of studies can pile up across an organisation, making it difficult to remember what was studied and where to find it. It's one of the reasons we've been thinking about AI in the context of Harmoni. ChatHarmoni gives researchers another way to work with the research they already have, asking questions of their data in natural language and going back into previous work when a new question comes up.

A current business question might have something relevant buried in a tracker from three years ago, an ad hoc project from last year and a segmentation study owned by another team. Finding those connections used to involve a fair amount of digging.

Sometimes the data is already sitting there

Steven Hill of RAC showed how his team is using AI to analyse complaints records and call transcripts alongside its Voice of Member program. The work, presented in "Unwrapping the Gift," helps RAC identify recurring complaint drivers and risk signals and understand how widespread particular problems may be.

The complaints and transcripts already existed, and RAC found another way to use them.

Most organisations have similar material in customer comments, call transcripts, complaints, open-ended responses and previous research. The volume of text has made some of it difficult to analyse at scale. AI makes more of that material available for researchers to examine.

Faster analysis doesn't fix the wrong question

In "Mirror Mirror — Modern Marketing Has Changed. Has Brand Tracking?" Nick Palmer of House of Brand and Heba Habashy of Commonwealth Bank looked at whether established approaches to brand tracking still line up with the concepts and language marketers use today.

We can analyse research faster than we could a few years ago, search it more easily and process far more text. That doesn't help much if we're tracking something the business no longer cares about or asking a question that doesn't tell us what we need to know.

Matthew Jorgenson of Human8 and Alla Nock of Kimberly-Clark took up another issue in "Creating a Human Spark": consumer closeness and the risk of brands becoming harder to distinguish as more marketing is generated using the same technologies and sources of information.

Processing more information about people doesn't necessarily mean we know them better. Researchers still need the context behind what people say and do, including the circumstances, motivations and experiences that aren't always obvious in the data.

An answer can sound right and still need checking

Cassie Roma's keynote, "The Trust Gap," dealt with evidence, belief and storytelling. She talked about the growing distance between having evidence and getting people to believe it, and what that means for how researchers communicate their findings. She used personal anecdotes to bring key points to life; this presentation stuck with me.

AI adds another consideration. It can produce a polished, plausible answer very quickly, which puts more responsibility on the researcher to know where that answer came from. Can we trace it back to the underlying data? Which studies support it? Is there evidence elsewhere that points in a different direction?

This is even more significant when AI is working across years of organisational research. Being able to see the research behind it is what allows someone to check whether an answer holds up.

Todd Sampson closed the conference with "Brain Power," using his own experiences and examples to explore brain plasticity and our capacity to keep developing our cognitive abilities throughout life. After a day of talking about what AI can learn, analyse and produce, I quite liked finishing with the human brain.