Following the revelations about Cambridge Analytica, much has been written about ‘psychographics’, one of the techniques the company is known to employ. There’s a lot of misunderstanding about psychographics: what it is, if it works, and whether Cambridge Analytica used it during the Trump campaign. Here’s a short primer.
First off: it’s a technique of targeting users with messages informed by their personality type. It mostly revolves around working out individual personalities based on the well-known ‘OCEAN’ domains (‘Openness, Conscientiousness, Extraversion, Agreeableness & Neuroticism). This is a pretty standard test used in personality testing. For decades psychologists have developed techniques to work out someone’s personality through questionnaires.
More recently, academics have set up these personality tests and posted them on Facebook, inviting people to respond. By cross referencing people’s survey answers against their Facebook Likes, they have tried to work out the correlation between the two and create algorithms that can determine from likes alone intimate details of millions of other users who hadn’t taken the survey. In 2013 for example, Michal Kosinski, one of the pioneers of this approach, co-published a study showing that easily accessible digital records of behaviour can be used to quickly and accurately predict sexual orientation, ethnicity, religious and political views, personality traits, intelligence, happiness, use of addictive substances, parental separation, age, and gender.
It is believed that Cambridge Analytica created personality profiles through this means. These profiles were then used to inform the style of messaging for each person, driven by each person’s OCEAN score. So let’s say you score highly on neuroticism, you’ll get a message which emphasises dangers, and so on. It’s all very emotional driven, which is one reason it feels quite manipulative.
Whether any of this actually works however is highly contentious. According to a 2017 study by Michal Kosinski, such techniques can attract up to 40% more clicks & up to 50% more purchases. Although bear in mind: a 50 per cent increase in clicks in digital advertising meaning 1 click per 1,000 becomes 1.5 clicks per 1,000. It sounds a lot in percentage terms but the overall levels of effectiveness are quite low.
To my knowledge, there are no studies about whether this works for voting behaviour. I suspect lots of the claims made by i.e. Cambridge Analytica about the effectiveness of this technique is salesperson bluster. I don’t think millions of minds can be manipulated by this kind of psychographic messaging. But, it will improve, and I am pretty sure it will be excitedly pursued in commercial advertising, and therefore future elections too. We should certainly be alert to its future evolution.
Sign up for our FREE Reaction Weekend Email
Read the week's best-read articles on politics, business and geopolitics
Receive offers and exclusive invites
Plus uplifting cultural commentary
A quite separate question is whether Cambridge Analytica actually used this technique. It is something the company has repeatedly boasted is its USP. We know they used psychographic techniques when they worked for the Ted Cruz campaign (they switched to Trump after Cruz was knocked out). But whether this helped the Cruz campaign or not is disputed, including by people on the campaign.
Brad Parscale (who ran Trump’s digital operations) denied they used psychographics for Trump. I interviewed Cambridge CEO Alexander Nix about this last year my BBC2 series ‘The Secrets of Silicon Valley’, and he also denied it, claiming they didn’t have time.But Nix did admit they took some ‘legacy data’ from the Cruz work they did and put them into the models they used for Trump.
In my opinion Cambridge Analytica’s more significant role for Trump was in building ‘universes’ of ‘persuadable voters’ (e.g. American moms worried about child-care who hadn’t voted before.) Creative types then designed specific ads for these universes, based on the specific things they cared about. It’s almost impossible to ever know what precisely made a difference when it comes to campaigning and advertising
It’s also said that Obama used this stuff too. This is true. But ad-tech and targeting has improved a lot (note Facebook partnerships with Axciom) since 2012. And bear in mind that Trump won Pennsylvania by 44 thousand votes out of 6 million cast, Wisconsin by 22 thousand & Michigan by 11 thousand. It was Cambridge Analytica (with the RNC) universe analysis that persuaded Brad Parscale to target these states extensively, against received wisdom.
In sum, the importance of psychographics as a technique is probably overhyped. But don’t let the focus on this dubious method obscure bigger challenges recent events highlight: of micro-targeting, data use & our outdated analogue election law.