Mia Suh UX & HCI Researcher [ Home Page ]
Collaborative Dynamic Queries : Supporting Distributed Small Group Decision-making



MOTIVATION
We often use technology individually to make a decisions for groups. We search Google Maps to decide where to hang out with friends, we search Airbnb to decide where to stay for family trip. As these decisions are not only for myself, it is critical to communicate with those to make the best decision that works for everyone. However, finding consensus among groups of people could be challenging. It may take such long time to communicate with everyone or to compromise different preferences among individuals. Even it is possible that not everyone is extroverted to actively express own preference. How could technology support these small-group decision making process?


GOAL OF PROJECT
In this study, we explored how to better support small-group decisions. Specifically, our goals were to :

  • Design a novel artifact that can increase group awareness in group decision-making
  • Evaluate the efficacy of the design artifacts in the controlled setting
  • Understand the pros and cons among each design alternatives for group decision-making

  • DESIGN OF AN ARTIFACT

    To help a group to reach a consensus on collaborative information seeking and decision-making, we designed Collaborative Dynamic Queries (C-DQ). C-DQ presents group awareness, which allows group members to be aware of (1) other members' filter selection ranges and (2) which decision candidates are within whose filter ranges. Showing group awareness helps a group to track others' preferences every moment and helps each member in the group to decide which action to take to reach a consensus. The figure below shows how C-DQ presents group awareness (in the right). First, C-DQ uses color encoding to distingish each member's identification (see "Member identification desing"). Horizontal bars in "UI widget design" present everyone's preference of filter ranges. Finally, vertical bars in "List design" shows information about which decision candidates are falling within whose filter preferences, which allows everyone to see who are agreeing or disagreeing with decision candidates.





    According to communication studies, a moderator plays various roles in group decision-making, such as :
    • Role 1. aggregating preferences over a set of decision criteria of a group and surfaces them to members.
    • Role 2. providing information about the decision candidates that match with individual and group preferences, and helps a group to identify if there is a feasible candidate that could lead to a decision.
    • Role 3. facilitating group consensus by identifying the sources of disagreement and suggests preferences that could lead to reach agreement within a group.
    We expected C-DQ would play a role of the moderator in group decision-making. Horizental bars in the left of the figure above takes the Role 1. Vertical bars in the right takes the Role 2. Finally, horizontal and vertical bars both take Role 3 and help a group to aware which particular directions they should change their mind to build a consensus. As a matter of fact, deploying a moderator in a decision-making loop imply the use of a trained human agent or an algorithm tailored to a specific domain. To date, deploying a moderator in everyday group decision-making has not been practical. Using C-DQ can be one solution for groups to make improved decision in everyday decision-making scenarios.




    Using this C-DQ, we designed and built a mobile web app which allows a group to search places around the Greater Seattle area. The web app incorporates five modules. At the top side of the screen, the app presents a list view module ("List" in the figure below) and a map view module ("Map"). A user can see more detail about each place in the list or the map module. Both modules present the same set of candidates in a different visual format and a user can toggle between a list or a map ("Toggle1"). At the bottom side of the screen, the app presents a C-DQ module ("C-DQ"), and a chatting module ("Chatting"). "Toggle2" in the figure below allows users to access either C-DQ or the chatting module. Division of the screen into top and bottom allows users to modify filter settings on the bottom while seeing the refined candidates on the top at the same time, which is an important principle for supporting an iterative information seeking process. Lastly, the app provides a search module ("Search").
    METHODS
    We used mixed-methods to evaluate the application with C-DQ. Specifically, we used the following methods:
    • Usability testing
    • Log-data analysis
    • Survey
    • Statistical analysis:descriptive statistic analysis, t-test, & a linear regression analysis with a nested term
    • Content analysis on chat-log and open-ended questions of survey
    • Qualitative analysis on interviews


    img

    The snapshots of the script for our usability study (left) and the survey questionnaires the participants responded after using the application (right)
    FINDINGS
    To examine the effect of C-DQ, we conducted a lab study consisting of usability testing as well as an controlled experiment with a within-subject design. 20 participants (in 5 groups) completed tasks using two different mobile apps (seeing or not seeing C-DQ). Results show that group's perceived efficiency, effectiveness, and the degree of satisfaction of decision-making process was significantly higher in the condition where they saw C-DQ. Along with this result, groups also made decisions with significantly less effort (e.g., wrote fewer chat lines, spent less time to make a decision) when they saw C-DQ. The more detailed result is well described in our paper below.
    PUBLICATION
    Sungsoo (Ray) Hong, Minhyang (Mia) Suh, Nathalie Henry Riche, Jooyoung Lee, Juho Kim, & Mark Zachry. (2018). Collaborative Dynamic Queries: Supporting Distributed Small Group Decision-making, In Proceedings of the ACM Conference Companion on Human-Computer Interaction (CHI'18). ACM. [PDF]


    MY CONTRIBUTION TO THE PROJECT
    • Co-lead the project
    • Design & Conduct evaluative studies
    • Lead statistical & qualitative analysis
    • Create the study materials
    • Communicate the results in the form of research paper and presentations to various audiences


    img img img