Personalization technology has the potential to optimize service for each person's unique needs and characteristics. One way to optimize service is to allow people to customize the service themselves; another is to proactively tailor services based on information provided by people or inferred from their past behaviors. These approaches function best when people know what they want and need, and when their behaviors and preferences remain consistent over context and time. However, people do not always know what they want or need, and their preferences often change. In addition, people cannot always articulate their preferences with the level of detail required for customization. The customized service that they want may be suboptimal for their needs. Finally, personalized services may become obsolete as people's preferences or contexts change, unless systems can detect these changes.

This thesis recasts personalization technology to accommodate uncertainties and changes in people's preferences and goals. I study personal service providers as a model for adaptive personalization that helps people customize their services and that adjusts service according to changes in people's preferences and goals. I derive design strategies for adaptive personalization, two of which I empirically evaluate.

The first strategy adapts service interaction styles to support long-term service usage. The first two studies investigate ways to detect people's preferred interaction styles with a robotic service – whether people treat the system as a relational being or a utilitarian tool – and the efficacy of personalizing service interaction based on this interaction preference. The next study explores how the relational interactions of technology service should be personalized over time in the context of a robotic snack delivery service in a workplace. Two types of adaptive relational interaction are investigated in a longitudinal field experiment – a social interaction strategy that adapts its conversation topic to knowledge common to an organization, and a personalized interaction strategy that learns about people over time and adapts its interactions accordingly. The results suggest that social and personalized strategies collectively improve people's cooperation, rapport, and engagement with the service over time; the strategies also influenced social dynamics in the workplace, facilitating the adoption of a robot into an organization.

The second strategy helps people gain insight into their needs and goals when they personalize service offerings. This strategy promotes reflection, helping people think through and articulate their needs and goals. I investigate different design variables for implementing a reflective strategy for technology service. I empirically evaluate its efficacy in the context of Fitbit, a physical activity monitoring service.

This thesis makes contributions to HCI, HRI, and interaction and service design. It broadens the concept of personalization discussed in HCI and HRI; designs and evaluates adaptive personalization strategies that accommodate uncertainties and changes in people’s preferences; draws attention to the dynamic nature of people's orientations to interactive technologies; and captures the humancentered design process of creating and implementing a robotic service.

232 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu