Truckee Meadows Community College
Mentors: David Feil-Seifer, University of Nevada, Reno
Rod Sidwell, Truckee Meadows Community College
Analysis of Human Interaction
The purpose of this work is to collect video recordings of people doing interactive two-person tasks, which require feedback. We are especially concerned with repeated instances of the same feedback. These data will then be coded to develop a generative statistical model of verbal and non-verbal behavior occurring during such feedback. This model will then be used in a future study (not included in this IRB) to generate realistic robot feedback behavior for interactive tasks.
One of our chief areas of interest the team wishes to study is human-human interaction in the general population. This is an important area to study as it will yield a baseline set of data to compare data obtained in other domains too. The tasks the team will ask the participants to perform include:
* Assembling a tan-gram puzzle
* Assembling a lego or erector-set structure
* Playing a game of simon-says
* Playing the card-game solitaire.
We have chosen the assembly tasks as an analogue to activities done in childhood related to cognitive development and problem-solving. It is likely that feedback generated during these tasks would transfer to feedback given during activities for child-robot interaction such as therapeutic interventions for children with autism spectrum disorders.
However, we foresee a much lower chance a requiring repeated feedback for these activities than with assistive activities. Therefore, we propose to introduce deception into our experiment design. We intend to have a single participant interact with an experiment confederate. We will tell the participant that the confederate is actually another participant. The confederate will intentionally make mistakes requiring the actual participant to provide repeated feedback.
We will be conducting research in a locked laboratory space. Video data will be collected showing the faces of participants. We will deceive participants by telling them that they are interacting with another participant. In fact,, they will be interacting with a confederate. This is needed to that one person participating in the task (the confederate) will intentionally make mistakes requiring the participant to provide repeated feedback. We will debrief participants after the data collection and inform them that the other participant was a confederate.
Each participant will perform one of the assembly tasks (tan-grams or erector-set) and one of the games (Simon-says, solitaire). Each task is expected to last between 5 and 10 minutes for a total of 10-20 minutes. We will record video of the feedback interaction between the participant and the confederate and code those data for verbal, gesture, and other nonverbal interaction behavior.