Be Transparent Now (Before You’re Made to be)
Off-the-shelf machine learning can give researchers great power. As any Spider-fan knows, there’s another side to that equation. Questions of great responsibility, morality, and ethics kept cropping up at IIeX.
In one of the week’s best talks, Andrew Konya of Remesh broke down the moral hazards of AI-powered research. Researchers handling large amounts of passive data will also find themselves handling tough metaphysical questions. When you can observe anything, how do you establish consent? When you can influence behavior effectively, what behavior is healthy? When you can train machines to look for truth, how do you define that?
A continual theme is transparency. Moving beyond mere compliance to be as plain-spoken and honest as you can be about what data you have and what your goal is in using it. Don’t bury the details in a “privacy policy” – be upfront about what you’re using AI to achieve. The alternative – as GDPR demonstrates – is the authorities deciding just how transparent you’re going to be.
No Data is an Island
As big brands start using AI methodology more, there’s been a shift in tone. AI is only as good as the data and model it uses – as System1 alumnus Alex Hunt of PRS InVivo put it, without good data your insights are artificial, not intelligent. But brands work with multi-source data, so the power of new technology for data synthesis is back on the agenda. Machine-assisted research needs to be broad, not just deep.
Some of the most impressive presentations were from companies like the UK’s Black Swan, pulling together multi-source data for analysis, and working out how to value each source to ensure machine analysis is spotting real patterns, not mirages.
One unintended consequence of all this is a revival of qual, as you need a strong source of baseline insights to begin asking the right questions – I don’t think I’ve ever seen so many excellent qual presentations at an IIeX event. Shout out to Labbrand and Problem Child for their brilliant ideas about ritual and post-truth research, respectively.