Introduction
This paper, presented as an extended abstract at IJCAI 2015, argues that the next frontier for intelligent systems lies in the "fusion" of personalization and persuasive technology. By tailoring persuasive strategies to individual user traits and using persuasive techniques to improve recommendation uptake, developers can create systems that are significantly more effective at changing behavior and improving user experience than either approach used in isolation.
The Problem: Siloed Approaches
The authors identify a missed opportunity in two maturing fields that have largely developed in parallel:
- Persuasive Technologies (PT): These aim to change user behavior (e.g., fitness apps, smoking cessation tools). However, most PTs use a "one-size-fits-all" approach, ignoring the fact that different people are moved by different types of influence.
- Personalized Technologies: Recommender systems are excellent at predicting user preferences but often struggle with the "cold start" problem (data acquisition) and getting users to actually act on recommendations (uptake).
The paper addresses this gap, proposing that their combination can solve these respective weaknesses.
Conceptual Framework for Fusion
As a position paper, this work synthesizes existing research into a conceptual framework for "fusion." The authors categorize this into two main directions:
- Personalization to Enhance Persuasion: This involves monitoring progress on parameters important to the individual, tailoring the language and timing of messages, and adapting the type of persuasion used (e.g., using "authority" for one user and "social proof" for another) based on individual susceptibility.
- Persuasion to Enhance Personalization: This involves using persuasive data acquisition (like rewards or simulated scenarios) to encourage users to provide rich data, and using persuasive explanations to build trust and increase the uptake of recommendations.
To implement this, the system can adopt various functional roles—acting as a tool, a medium, or a social actor—to better engage the user.
Key Insights: The Feedback Loop
A core insight of the paper is the "Susceptibility Gap": persuasiveness is not a property of the message alone, but a relationship between the message and the recipient's personality. Personalizing the strategy of persuasion is the "holy grail" of this fusion.
This creates a synergistic feedback loop. User modeling informs personalization, which improves the user experience. A better experience, in turn, encourages users to provide more information and feedback, which further refines the user model.
Evidence from Research
The paper summarizes findings from several successful "fused" studies:
- Smoking Cessation: A study showed that smokers receiving personalized feedback based on specific habits had significantly higher quit rates than those receiving generic information.
- Motivating Games: In physical activity games, adapting difficulty to a player's skill level stabilized the persuasive power, preventing boredom and abandonment.
- Argument Modeling: Research into modeling a user's position on specific topics proved that understanding baseline involvement is critical for selecting the most effective persuasive argument.
Limitations and Ethical Considerations
The authors acknowledge significant hurdles for this emerging field. Building models of "persuasion susceptibility" is difficult because it requires explicit, psychological data that is harder to infer than simple interests. Furthermore, measuring impact requires complex longitudinal studies rather than short-term metrics like click-through rates.
Finally, the paper warns of the ethical fine line between persuasion and manipulation. There is an inherent conflict between a service provider's commercial goals and the user's best interests, necessitating a balanced and transparent approach to how these technologies are deployed.
Summary updated