2016 Netgain – Emotion Analytics Can Predict What People Do and Explain Why They Do It

The following are my notes from the 2016 Net Gain presentation of “Emotion Analytics Can Predict What People Do and Explain Why They Do It”, by Lana Novikova of Heartbeat Ai Technologies Inc.. This has been posted shortly after the end of the presentation, so there has been little editing and hence there will be typos.

Had an idea 15 years to create a product to analyse open ends, while she was at another job when internet data collection was just starting. In March 2016 she had built a prototype of her product. Product would analyse phrases and anaylse the meaning into different emotions.

First project was based on how women would feel toward chocolate – feelings of anger, joy and sadness.

Last election – bubbles of joy toward feelings for Justin Trudean, “the opposite” toward Harper.

Emotion Analytics can predict what people do and why they do it.

Using this technology the company predict the Australian election very accurately, which polls said was quite close.

Daniel Kahneman: “Nobody would say, ‘I’m voting for this guy because he’s got the stronger chin,’ but that, in fact is partly what happens.”

Q: Where do we find the answer in the sea of data?

How do you get information from the human brain?

Language is the oldest human tech – according to Chomsky emerged between 60k and 200k years ago.

Customer service reps can tell from your tone in voice – within 30 seconds – how you are feeling and what type of intervention will be needed.

Use the following emotions: void, fear, anger, joy, trust, love, surprise, sickness, body sense

The software has 8,000 words and is coded to one of the emotions that they are using in the software.

For example: In a survey can ask: “How do you feel about ____ (Candidate)” — and then use the algorithim to segment each response into emotions.

Australian election polling

  • Used Google Consume Surveys
  • N=943
  • May-July 2016
  • 5 questions
  • Results of election Turnbull 50.36%, Shorten 49.64% – margin of 0.72%
  • Rational model component — Voter Preference + Voter expectations of who woudl be PM
  • Example: Q: How do you feel about Shorten/Turnbell becoming the next PM? Responses of emotional breakdown were pretty much the same between each candidate.

Examples – banking study found people do feel emotional toward their bank, but customers do not toward telecom companies

 

Leave a Reply