Smart, Context-Aware, Peer-to-Peer Service and transaction Matching (2011-2018)

Funded by the National Science Foundation (NSF), with Penn State U. and CMU, I am working on developing a system to match service providers and receivers in real-time, based on both personal profile and dynamic context. My main collaborator thus far has been Prof. John M. Carroll, otherwise known as Jack. Other collaborators on this project are many; Prof. Anind Dey and his students, Afsaneh Doryab and Alaaeddine Yousfi, at CMU have worked on the context-aware engine. Simon Tucker worked on the semantic graph for user modeling. Sara Cambridge, Alexander Ambard, Daniel Turner, Kamila Demkova, Christina Gossman and Stephanie Snipes worked as my field research and design assistants. Prof. Carroll's student Kyungsik Han developed a mobile app (not shown here) and post doc, Patrick Shih worked on survey analysis (also not shown here). 

Working closely with Anind and his students, who have plenty of expertise in contextual intelligence, I am defining the way in which data can be used to make inferences about the best people to perform a task for someone else (could be another person, a business or some other entity).

For example, a user’s predicted travel path can be used as a source of data for making an inference such as calculating the user with the shortest predicted course deviation required to pick up someone from the airport and drop them at home.

The screenshot to the right shows a temporary interface the CMU team used to test request matching. It's not very intelligible to non-CMU CS PhD.s, but we found it useful internally.

By the way, the output from NSF-funded research is open source and we are looking for customers, both for- and non-profit that will be interested in our technology.

Screenshot of current prototype web UI for transaction matching showing a match for a ride from the airport. The system recommends the best person to fulfill the request, based on predicted travel pattern.

Motivations for Participation in Peer-to-Peer Exchanges

Part of our effort is devoted to uncovering and analyzing user and other stakeholder needs and motivations for participation in peer-to-peer transactions and designing our mobile service exchange solution to tap into these. Based on past theoretical work by scientists such as Robert Trivers, Icek Ajzen, Edward L. Deci and Richard M. Ryan who have investigated the psychology of human motivations, I developed a motivational analysis framework. My research assistants and I are using this framework for classifying interview data from users as receivers (URs) and providers (PRs) of peer services and from service providers talking about their claims about what motivates their users and what aspects of motivation they designed their systems to leverage.

My research assistants, Alexander Ambard, Dan Turner, Christina Gossmann and Kamila Demkova classified talk in interview transcripts according to the framework on the left here. Each cell in the table represents frequency of mention of a motivation as a percentage of the highest count of any motivation in the column for each stakeholder.

You don't have to be a genius to notice that there is something peculiar going on here. Providers are talking a lot about users being motivated by idealistic motivations around creating a better community or society and they talk a lot about designing their systems to encourage this. However, users, both as providers and recipients of goods and services are much more motivated by self-interested concerns such as getting a service or thing, getting good value or being paid for something.

Interestingly everyone seems to think that social motivations are important.

If you break out the profits and the non-profits, you still see mismatches, but for the for-profits, the social motivations seem to be less important than for the non-profits, even though service providers do talk a lot about how they design for these.