Shared Quests, Not Individual Points or No Game Layer: Group Goal Pools in AI-Coached Community Walking and the Members Who Fall Behind

Abstract

In a 16-week community walking program, 48 groups of five or six members (276 people) used the same AI coach and were randomly assigned by group to one of three game layers. In "shared-quest" groups (16 groups, 93 members), the coach set a weekly distance target for the group as a whole, showed only the group's combined progress, and suggested that members walking ahead of their usual pace invite another member along. In "individual-points" groups (16 groups, 92 members), each member earned points for their own distance, visible only to themselves. "No-game" groups (16 groups, 91 members) received the coach's walking suggestions alone. In weeks 13 to 16, 62% of shared-quest members still logged at least three walks of 20 minutes or more every week, against 51% of individual-points members (a 22% gap) and 49% of no-game members (a 26% gap). Among members in the least active third of their design at baseline, the figures were 48%, 39% (a 25% gap), and 40% (a 21% gap). Consenting shared-quest members logged 1.9 joint walks with another group member a month, against 1.5 for individual-points members and 1.6 for no-game members. Members in the most active third of the shared-quest design walked 6% less than their counterparts in individual-points groups, and 8 of 31 reported at least once feeling they were carrying the group.

1. Introduction

A shared goal works by making one member's walk count toward everyone's target. The game layers most common in coaching apps do the opposite: points, badges, and levels belong to one person and record what that person did. Even without a leaderboard, which an earlier issue of this journal examined in AI-coached skill practice, an individual score frames each member's activity as their own account.

In a group of neighbors who walk together, that framing leaves out what keeps a walking group going, which is that members turn up for each other. We asked whether an AI coach that counted the group's progress, rather than each member's, would keep more members walking, and in particular whether it would keep the members least likely to continue.

2. Method (Summary)

The program was organized by community centers in four towns, which formed 48 walking groups of five or six adults from people who signed up together or were matched by neighborhood. All 276 members used the same AI coach on their phones, which suggested routes and weekly walking plans and logged walks through the phone's step counter. Groups, not individuals, were randomly assigned to a game layer. Weeks 1 and 2 were a run-in period in which all three designs received only the coach's route and planning suggestions; the assigned game layers ran from week 3 to week 16.

In shared-quest groups (16 groups, 93 members), the coach set a weekly distance target for the group based on its run-in activity in weeks 1 and 2 and showed a single progress bar toward it. No individual's contribution was displayed. When a member was walking ahead of their own usual weekly pace, the coach suggested they invite another member on their next walk. In individual-points groups (16 groups, 92 members), each member earned points for their own logged distance and could see their points and a personal weekly target, but not other members' points. In no-game groups (16 groups, 91 members), the coach gave the same route and planning suggestions with no points or targets.

Members were ranked by logged distance in weeks 1 and 2 against all other members assigned to the same design, and the least active and most active thirds of each design were identified (31, 31, and 30 members in the least active thirds; 31, 31, and 30 in the most active). The primary outcome, from step-counter logs, was whether a member logged at least three walks of 20 minutes or more in each week from 13 to 16. A joint walk was counted when two members' logs from the same group overlapped in time and location for at least 15 minutes and both members had agreed at enrollment to location matching, which 84 of 93 shared-quest, 81 of 92 individual-points, and 82 of 91 no-game members did. Joint-walk figures are means per consenting member. At week 16, all members completed a short survey on their experience of the program; shared-quest members were also asked whether they had felt at any point that they were "carrying" other members toward the group target.

3. Findings

Shared-quest members kept walking at higher rates. In weeks 13 to 16, 58 of 93 shared-quest members (62%) logged at least three walks of 20 minutes or more every week, against 47 of 92 individual-points members (51%, a 22% gap) and 45 of 91 no-game members (49%, a 26% gap). Individual points and no game layer were close to each other.

The pattern held among members who started in the least active third of their design: 15 of 31 in shared-quest groups (48%) were still walking at that level, against 12 of 31 in individual-points groups (39%, a 25% gap) and 12 of 30 in no-game groups (40%, a 21% gap).

Over weeks 3 to 16, consenting shared-quest members logged a mean 1.9 joint walks with another member of their group a month, against 1.5 for individual-points members (a 27% gap) and 1.6 for no-game members (a 19% gap). Among consenting members of the least active thirds, the figures were 1.6 in shared-quest groups and 1.3 in both other designs.

The cost fell on the most active members. In shared-quest groups, the most active third walked a mean 24.1 km a week in weeks 13 to 16, against 25.7 km for the most active third of individual-points groups (6% less) and 25.2 km for that of no-game groups (4% less). In the week-16 survey, 8 of the 31 most active shared-quest members (26%) said they had at least once felt they were carrying the group.

4. Discussion

The mechanism we expected was the invitation, and the joint-walk figures are consistent with it, though we did not record who invited whom. When the coach counted the group's distance rather than each member's, a member walking well had a reason to bring someone else along, and the least active members joined somewhat more joint walks in shared-quest groups than elsewhere. A walk with a slower neighbor counted for the group just as a fast solo walk did, so walking with a slower neighbor cost the stronger walker nothing in the group's score, only some of their own pace.

Who benefited matters more than the averages here. The benefit appeared to reach the least active third about as much as it reached the group as a whole, which incentives aimed at individuals do not always achieve. Individual points did little for anyone compared with no game at all.

The most active members paid for this. They walked somewhat less than the most active members of the other two designs, and about a quarter of them felt, at least once, that they were carrying others. The coach never showed who had contributed what, which may have kept that feeling from turning into resentment, though the design cannot tell us. A shared target asks the strongest members to share their walking with the others.

5. Limitations and Future Directions

Members knew their group's design, and groups that formed from people who signed up together may have differed in cohesion from groups matched by neighborhood; assignment by group balanced the two only on average across 48 groups. Joint walks were counted only for members who agreed to location matching, and walks together without that consent were missed. The "carrying" question was asked only of shared-quest members, so we cannot say how often active members in other designs felt the same, and it was asked once, at the end, so it does not show when the feeling arose or whether it passed. The least active thirds held only 30 or 31 members per design, so the subgroup differences rest on three members each and are imprecise. Step counters record distance, not effort, so a slower member's walk may have been harder than its distance shows. A follow-up will test whether showing members, privately, how many joint walks they started changes how the most active members feel about the shared target.

Editorial Note

This report underwent CfAHC's Rapid Review Protocol, including AI-assisted literature cross-checking and a two-reviewer human sign-off.