Measuring Personality and It's Use in Advertising

The text and images below are drawn from a presentation Dr. Wu gave at Northwestern University. You can view the video online (section excerpted below is from 34:00-38:00) but the audio is a bit fuzzy, so it might be easier to read the text below.

"I just published a paper looking at similarity in personality between romantic partners. So I measured a couple's personality from their Facebook likes and then compared the similarity between them. This would be an example of studying personality in the context of close relationships using this measure instead of self-report. And this question has been studied using self-report for a long time, but we've identified specific biases and show that not only can our method replace self-report in studying this question, but also has a lot of advantages that were not covered by self-report.

Another example would be a research piece done by my colleague Sandra Matz. So she was looking at tailored advertising based on psychological traits. Using our models she identified people of high extroversion versus low extroversion and she created and tailored commercial content for these two different groups of people. Sort of like putting people into bins, high versus low extroversion."

"And the results were amazing. Those personality optimized advertisements were shared three times more often than a generic one. So it really shows that people genuinely enjoyed the ads with content that are better suited for their personality. And these results of course have a lot of commercial implications. Like, in the past ads targeting based on psychological traits has not been very feasible. Because it's impossible for marketing companies to ask all the potential consumers to do a questionnaire and know their personality. So most of the ads target easily observable attributes like demographics. But with this kind of technique we can now infer psychological traits from existing data to help companies better market their products."

"Here's another example of a tailored ads targeting campaign that my previous lab has done with Hilton. They were creating differential content for people of high extraversion versus high agreeableness -- not versus, but different groups of people -- and for high extroversion people they were talking about having fun, like partying, but for high agreeable people they were talking about family relationships.

So far I've been talking about predicting personality from one type of digital behavior -- Facebook Likes -- and there are definitely many other options. My colleagues and collaborators have been playing with a few other types of digital data and I want to show you an example of those. "

"Here we can assess personality from the language people use in their Facebook statuses. As many of you know, status updates are about posting your feelings, thoughts, and life events. Here are two word clouds showing and contrasting extroverts versus introverts choice of words. Extroverts like to talk about party, weekend, girls, and introverts on the other hand like to talk about the internet, alone stuff, and computers. [laughter]"

"We also were able to assess personality from people's profile pictures. I love these images of aggregated faces showing the most introverted people, with glasses, and the most extroverted people, with makeup."