ML and Behavioral Economics for Personalized Choice Architecture
Data Science Academic ResearchMachine LearningModelingbehavioral economicsMachine Learningpersonalized choice architectureposted by ODSC Team September 19, 2019 ODSC Team
Personalized choice architecture is a new and interesting field surrounding the idea that we can predict and influence consumers’ choices, ensuring that they get the best product or best outcome possible. In a recent paper by Emir Hrnjic and Nikodem Tomczak (researchers at NUS Business School and the National University of Singapore, respectively) machine learning and behavioral economics have been brought together to attempt to manifest and perfect this idea of personalized choice architecture by creating so-called “nudges.” The paper discusses what is involved in these “nudges,” looks at some of the potential use cases for this technology, some of the threats of it, and suggests further research in adversarial attacks to try and prevent the misuse of it.
[Related Article: Best Deep Reinforcement Learning Research of 2019 So Far]
What is a nudge?
In this case, a nudge is a means to encourage consumers to do a specific task or buy a specific product. Think of the push notifications from your apps, or a reminder of what’s in your online shopping cart. These are simple versions of nudges, intended to influence your behavior.
In this case, a nudge is a machine learning program which has been trained using theories of behavioral economics—a field dedicated to learning about what influences and shapes consumers’ behavior—in order to better predict what consumers will want on a more individualized basis. Where in the past nudges have been deployed to mass markets or on a timer-type system, these new programs could create individualized suggestions.
Some Potential Uses
While the obvious place for this technology to be used is in online shopping, there are actually multiple different situations in which these nudges could be beneficial. The paper goes into more detail on each of these cases, but some of them are:
- Corporate policy: nudge employees when they need to get a vaccination or fill out some paperwork, and communicate in a way that they’ll actually respond to
- Healthcare: encourage people to follow their post-appointment directions
- Conservation: nudge homeowners or businesses if they’re overusing/misusing resources
Like any new technology, personalized choice architecture comes along with some potential issues. First and foremost, there’s the potential for the technology to be misused or attacked by adversarial networks. Further research has been encouraged to better understand how to prevent these.
[Related Article: 10 Compelling Machine Learning Dissertations from Ph.D. Students]
More specifically, further research will need to be done to better understand when nudges are helpful and when they could actually be counterproductive. Are there any situations where a nudge makes people less likely to participate, aka where no nudge would have been better? It will take time to understand these issues completely, but for now, it’s exciting technology that could do a world of good.