P3RD
← Observatory

Research in progress

The value of contribution

How might people find collaborators through the work they share and the trust they build?

A review of twelve published sources on how shared work becomes trust between strangers, followed by a method for a small study Project 3rd Rock has not yet run. Each claim links to its source.

Summary

What the evidence says so far

The nearest evidence comes from software, mostly GitHub. There, shared work does act as a signal, and the signal is uneven. No study reviewed here measured how collaborators find each other in climate, civic, or learning communities, which is the gap the proposed method addresses.

  • Half of contributors in GitHub's 2017 survey said their open source work was somewhat or very important in getting their current role.

    GitHub, Open Source Survey 2017

  • People look up strangers' activity histories and form impressions from them, and those impressions shape how receptive they are to a contribution.

    Marlow, Dabbish and Herbsleb, 2013

  • Project maintainers weigh the social connection to a submitter alongside the technical quality of the work when deciding whether to accept it.

    Tsay, Dabbish and Herbsleb, 2014

  • Women were more likely than men to seek help from people they already knew rather than from strangers in public (22 percent against 6 percent), and 21 percent of people who saw a negative interaction stopped contributing to a project.

    GitHub, Open Source Survey 2017

The question

Much of the research on collaboration asks what makes a team work once it exists. This entry asks the earlier question: how two people who have not met come to believe that working together is worth trying, and what the work they have already shared does to make that belief possible.

Why it matters

Climate, civic, and learning work depends on people finding the right collaborators without a hiring process or a shared employer. In practice that often happens through visible work and through introductions. If the signals people rely on are narrow, uneven, or easy to misread, capable people are overlooked and trust is given for the wrong reasons.

50% of contributors

Fig. 01. Said open source work was somewhat or very important in getting their current role. Source: GitHub, Open Source Survey 2017 ↗. Self-reported; 5,500 randomly sampled GitHub respondents plus a non-random offsite sample. Software contributors only.

One reading of the evidence

Read together, three studies of GitHub describe a sequence worth testing: work leaves traces, traces form impressions, and impressions inform whether to collaborate.

  1. Stage 01

    Contribution trail

    Shared work leaves visible traces that others can read, such as edits, reviews, and replies.

    Dabbish et al., 2012 ↗

  2. Stage 02

    Impression and trust

    People look up the history of someone unfamiliar, and form an impression that shapes how receptive they are.

    Marlow et al., 2013 ↗

  3. Stage 03

    Collaboration

    Evaluation of a contribution draws on technical quality and on the social connection to the person submitting it.

    Tsay et al., 2014 ↗
Fig. 02. A concept diagram, not a measurement. Each stage restates a finding from one cited GitHub study. The sequence is an organising reading of those studies, and no study tested it end to end.

What existing evidence says

Most of the empirical work concerns software. It is the nearest evidence available. Twelve sources follow, grouped by theme. Each card gives the claim, its source, and what it cannot show.

Signals and reputation

  • Job-market signaling theory describes how observable choices and credentials can convey otherwise hidden qualities to someone deciding whether to work with a person.

    Job Market Signaling ↗ Michael Spence. The Quarterly Journal of Economics, MIT Press. August 1973. Checked 2026-10-08.

    Limitation: A theory of labor markets and education. It does not describe public work portfolios or volunteer collaboration, and it was not tested against them here.

  • Lerner and Tirole argue that the economics literature on career concerns, together with industrial organization theory, can explain features of open source projects, including participation that builds reputation for later labor-market rewards.

    Some Simple Economics of Open Source ↗ Josh Lerner and Jean Tirole. The Journal of Industrial Economics, vol. 50, no. 2, pp. 197-234. June 2002. Checked 2026-10-08.

    Limitation: Verified from the bibliographic record and abstract only. The full text was not read, so details of the argument are not restated here.

  • In GitHub's 2017 Open Source Survey, half of contributors said their open source work was somewhat or very important in getting their current role.

    Open Source Survey 2017 ↗ GitHub, with offsite recruitment help from the Open Source Initiative. GitHub. 2017. Checked 2026-10-08.

    Limitation: Self-reported, from 5,500 randomly sampled respondents across more than 3,800 GitHub repositories plus a non-random offsite sample of over 500. It reflects software contributors in 2017, not climate or civic communities, and the survey page does not describe how people found collaborators.

Weak ties and discovery

  • Granovetter argued that weak ties link otherwise separate clusters of people and so carry information across them.

    The Strength of Weak Ties ↗ Mark S. Granovetter. American Journal of Sociology, vol. 78, no. 6, pp. 1360-1380, University of Chicago Press. May 1973. Checked 2026-10-08.

    Limitation: Verified from bibliographic records and secondary summaries; the full text was not read. The paper concerns information flow and job finding, not project collaboration.

  • A randomized study on LinkedIn, covering over 20 million people across five years, found that weak ties help job mobility up to a point and then level off, in an inverted-U pattern. Strong ties mattered more in less digital industries.

    A causal test of the strength of weak ties ↗ Karthik Rajkumar, Guillaume Saint-Jacques, Iavor Bojinov, Erik Brynjolfsson, and Sinan Aral. Science, vol. 377, issue 6612, pp. 1304-1310. 15 September 2022. Checked 2026-10-08.

    Limitation: Measures job transmission on one professional network, not collaboration on shared work. Read from the Stanford Digital Economy Lab listing and a search summary, not the full paper.

Trust and impressions

  • Interviews with central and peripheral GitHub users found that people draw a rich set of social inferences from visible activity, such as guessing a developer's goals from their edits, and turn these into strategies for coordinating work, building skills, and managing reputation.

    Social Coding in GitHub: Transparency and Collaboration in an Open Software Repository ↗ Laura Dabbish, Colleen Stuart, Jason Tsay, and Jim Herbsleb. ACM CSCW 2012, pp. 1277-1286. 2012. Checked 2026-10-08.

    Limitation: Qualitative interview study of one platform at one moment. Read from the abstract and a search summary.

  • A qualitative study of GitHub found that users look up information about one another, and that impressions built from activity history across projects and past collaboration with high-status projects shape how receptive they are to strangers' contributions.

    Impression Formation in Online Peer Production: Activity Traces and Personal Profiles in GitHub ↗ Jennifer Marlow, Laura Dabbish, and Jim Herbsleb. ACM CSCW 2013. February 2013. Checked 2026-10-08.

    Limitation: Qualitative and platform-specific. Describes how impressions form, not whether they predict good collaboration.

  • A study of pull requests found that project managers use both technical information and the strength of the social connection with the submitter when evaluating a contribution, and that well-established projects were more conservative in accepting them.

    Influence of Social and Technical Factors for Evaluating Contribution in GitHub ↗ Jason Tsay, Laura Dabbish, and James Herbsleb. ICSE 2014. 2014. Checked 2026-10-08.

    Limitation: Observational data from GitHub. Association, not proof that social ties cause acceptance. Read from an abstract summary.

  • In the same survey, women were more likely than men to seek help directly from people they already knew well rather than from strangers in a public forum (22% against 6%).

    Open Source Survey 2017 ↗ GitHub. GitHub. 2017. Checked 2026-10-08.

    Limitation: A single survey item about seeking help, not about choosing collaborators. It suggests prior familiarity can shape who people approach.

Newcomers and barriers

  • A conceptual model built from a literature review, contributor responses, student contributors, and interviews with 36 developers from 14 projects identified 58 barriers to a newcomer's first contribution, 13 of them social.

    Social Barriers Faced by Newcomers Placing Their First Contribution in Open Source Software Projects ↗ Igor Steinmacher, Tayana Conte, Marco Aurelio Gerosa, and David Redmiles. ACM CSCW 2015, pp. 1379-1392. 2015. Checked 2026-10-08.

    Limitation: Software projects only. The barrier counts describe the authors' model, not how common each barrier is.

  • In the same survey, 25% of women and 15% of men reported encountering language or content that made them feel unwelcome. Among people who had experienced or witnessed a negative interaction, 21% said they stopped contributing to a project because of it.

    Open Source Survey 2017 ↗ GitHub. GitHub. 2017. Checked 2026-10-08.

    Limitation: Self-reported, software contributors only. Shows that welcome affects whether contribution continues, not how collaborators find each other.

Peer production

  • OpenStreetMap has been studied as a case of volunteered geographic information, with analysis of how its contributors' activity is distributed over time.

    Analyzing the Contributor Activity of a Volunteered Geographic Information Project: The Case of OpenStreetMap ↗ Pascal Neis and Alexander Zipf. ISPRS International Journal of Geo-Information, vol. 1, no. 2, pp. 146-165. 2012. Checked 2026-10-08.

    Limitation: Only the bibliographic record was verified. No figures from the paper are used. It is listed as a model for contribution-trail analysis, not as evidence about trust.

What is unknown

The sources above leave the central question mostly open. None measures how people find collaborators in climate or civic communities, and the software evidence may not transfer.

  • Do the signals studied on code platforms carry over to climate, civic, and learning communities, where work is less often a public artifact?
  • Which traces do people trust, and which do they discount, when they decide whether to approach a stranger?
  • How much does an introduction through a mutual contact change the weight given to public work?
  • Who is excluded when visible output is the main signal, including people with less time, less confidence, or private work?
  • What does a newcomer need to see in order to believe a first contribution will be welcomed?
  • Can trust built through shared work be described honestly without reducing it to a score?

Proposed method

This is a proposal, not work already done. The aim is to learn from a small number of communities and to describe the limits plainly.

  1. Step 01

    Define a contribution trail

    Agree a small set of observable traces for a person's work: shared artifacts, reviews, replies, maintained resources, and credited collaboration. Record which traces each community actually exposes.

  2. Step 02

    Map trails in two or three communities

    Choose communities with different norms, for example a software project, a civic or climate data community, and a learning community. Describe how work becomes visible there, using public records and with permission where records are not public.

  3. Step 03

    Interview people about specific collaborations

    Ask each participant to recount one collaboration that began through shared work and one that did not. Ask what they saw first, what they checked, and what made them say yes or no. Interviews would be consented, summarised without identifying details, and shared back with participants.

  4. Step 04

    Compare what was seen with what was trusted

    Set the visible traces against the reasons people give. Note where the traces mislead, are missing, or favour people who already have time and standing.

  5. Step 05

    Publish findings with limits

    Report what was and was not observed, how many people were involved, and what remains uncertain. Invite corrections before treating any pattern as established.

Source list

Spot a source correction or a missing qualification? Email a correction for The value of contribution.