Difference between revisions of "Zooniverse research (2016-2017 archive)"

From CommunityData
Line 128: Line 128:
[https://docs.google.com/document/d/11K-clLkPeR4qIp3XvwCdobQo5QkmQXKDBwN2HB530_A/edit?usp=sharing List of Dependent Variables]
[https://docs.google.com/document/d/11K-clLkPeR4qIp3XvwCdobQo5QkmQXKDBwN2HB530_A/edit?usp=sharing List of Dependent Variables]
[https://api.zooniverse.org/projects/serengit/status API endpoint: https://api.zooniverse.org/projects/<projecy_name>/status]

Revision as of 20:31, 28 July 2016

Playground for Jim, Darren, and Aaron to build up a project analyzing volunteer management in the Zooniverse.

Action items for next meeting

  • Jim: Create a typology of different types of information seeking (with examples)
  • Jim: Read Tevvan papers on social search
  • Aaron: Database access?

Pending Action Items

  • Develop list of models/processes we think are happening.
  • Develop list of measures/variables we want.
  • More reading (start w Zooniverse and shared leadership then move on from there).
    • almost wikipedia - Mako
    • papers on stack exchange, leadership through expertise, leadership through knowledge sharing
    • review open-source software motivation/leadership papers (Kevin Crowston lit review 2008 "free/libre open source software development, what we know and don't know")
    • CCI review chapter in zotero
    • volunteer engagement
    • (shared) leadership
    • community management
    • citizen science
    • Kevin Crowston
    • Carsten Oesterlund
    • Andrea Wiggins
    • Rob Simpson
    • Chris Lintott (PI of Zooniverse)

Meeting logs


JM DG & AS narrowed research focus to project outcomes associated with information seeking, specifically tracking users who have engaged in information seeking and judging performance before and after.


Jim circulated the BYOR Planning Document and typology of leadership and volunteer mgmt practices, and concluded that "shared leadership" might not be the best conceptual framing. The group brainstormed updated RQs and methods of analysis.


Darren joined the meeting/project. Discussed more observations, recent readings, and responses to BYOR feedback. Set goals to build a leadership typology and plan revisions to planning document.


Discussed "triangulation" (theory, observations, folk-knowledge). Discussed observations from community talk and best practices and connections to literature. Created list of other relevant theory and literature.


Talked about predicted probabilities for unrelated projects. Revised and updated work goals for next week.


Laid out plans for the quarter. Set some preliminary work goals.


Spring 2016

  • Fully developed planning document including a thorough analysis plan.
  • A cleaned up dataset (variables and measures that we can model; not just raw database records).

Preliminary Project Overview

Understanding Effective Leadership and Volunteer Engagement in Citizen Science


Prior research investigating the mechanisms of effective crowd work and citizen science has focused on volunteer/worker motivations, algorithmic and computational techniques for quality control/improvement, and task/incentive design. Despite a number of findings indicating that the patterns of interaction between requesters and volunteers/workers shape participant experiences and work contributions in important ways, these interactions remain largely unexplored as a domain of analysis, design, and intervention. This is arguably an especially important design space in the context of citizen science, where "the crowd" consists entirely of volunteers and poor volunteer management or relations would likely result in complete project failure.

As task requesters in citizen science are scientists who perform important leadership functions for the project as a whole, we draw on prior literature analyzing effective leadership in volunteer online communities. Several studies by Haiyi Zhu and colleagues in the context of peer production show that "shared leadership" behaviors increase newcomer motivation and participation along multiple dimensions, but may adversely affect experienced participant motivation and participation.

This study will contribute a comparative analysis of requester-volunteer interactions in a sample of citizen science projects from within the Zooniverse platform. Building on literature on crowd work, citizen science, and studies of effective leadership in online communities and peer production, we construct models to test whether shared leadership behaviors lead to enhanced citizen scientist engagement.

Analytic approach:

Use observational data drawn from Zooniverse database records to build variables and multilevel models that test whether science team member behaviors and science team - volunteer interactions consistent with shared leadership predict increased volunteer engagement and higher quality contributions. <more details needed here>

Dependent variables (all nested within projects):

  • Count of volunteer contributions
  • Median/total quality of contributions (using gold-standard data where available)
  • Volunteer survival rate: time on zooniverse platform, completed project (0/1)
  • Number of active volunteers throughout project?
  • Risk of highly negative interactions between volunteers and science team members (if we can identify these)
    • from my observations I have yet to find these.
    • these seems like an independent, not a dependent variable

Dimensions of Analysis


shared leadership (from Zhu et al.)

  • transactional (positive feedback)
  • aversive (negative feedback)
  • directive (directive message)
  • person focused (social message)

Group Identity (from Ren et al.)

  • identity attachment to group within the community
  • bond based attachment to individual members
  • attachment to the large community

Network Exchange Patters (from Faraj & Johnson)

  • direct reciporcity
  • indirect reciprocity
  • preferential attachment

social Q&A?


  • building a great project
    • built in cooperation with volunteers.
    • discuss your research goals (especially on project page)
    • two way communication--volunteer testing
  • the launch rush
    • Write a newsletter.
    • have a promotion plan, recruit from outside zooniverse
    • use talk during initial launch
    • moderators for talk
  • in for the long hall
    • reminder emails
    • create community--more commitment from volunteers involved in talk
    • positive feedback, report on accomplishments


  • information seeking

Data and measures we want

  • Users for every project we study
    • Account creation date / project
    • Projects active in
    • Tags for various roles (participant, moderator, science team, zooniverse employee)
      • Dates roles acquired/dropped

Resources & Links


List of Dependent Variables API endpoint: https://api.zooniverse.org/projects/<projecy_name>/status


BYOR Planning Document

typology of leadership and volunteer mgmt practices


Zooniverse Best Practices


04-04-16 to 04-19-16

  • we should split between volunteer posts and scientist posts to forums
  • volunteers and scientists use the forums for different purposes
  • "folk-knowledge" only applies to scientists/admins
  • jungle rhythms uses talk to engage community, single researcher/moderator?
    • use of talk is somewhat task dependent?
    • small community, but still no social leadership
    • all the same researcher as point of contact? creates community? form of social leadership?
  • different communities have different methods of linking researchers to volunteers
    • Shakespeare world has explicit thread in forum
    • Some use blogs

BYOR 04-18-16

  • do a better job explaining Zooniverse (the platform), the roles within the community, and the types of interaction on the community forums
  • refine research questions
    • first question is causal but the research design doesn't support causal inference
    • second question can't be answered by independent variables as currently defined. Consider using qualitative methods and/or NLP
    • define "success" in concrete terms (success of the project vs success of the interaction), success is actually a collective outcome
    • define measures of quality
    • control for individual level variables (one of the
  • relevant lit
    • community management
    • kickerstarter
  • Aaron's notes (many redundant)
    • Read syracuse folks
    • Talk to Adler
    • What is it? Explain Zooniverse a bit more.
    • Explain measures
    • Explain puzzle/interest x2
    • Explain how you are gathering qualitative data
    • First question seems to require causal identification
    • Second question seems like it needs text analysis
    • What does this communication look like? What roles do volunteers take? x2
    • Community mgmt literature?
    • Refine RQs
    • Explain variables in relation to invididual level (vs conversation level?)

04-04-16 to 04-13-16

Zooniverse Best Practices

  • building a great project
    • Some of our most successful projects in terms of engagement have been those built in cooperation with volunteers. (all)
    • You can (and should) discuss your research goals in depth on your project’s Research page. (directive)
    • two way communication--volunteer testing (including the formal Zooniverse "project review" stage) will reveal things to alter (directive, social)
  • the launch rush
    • importance of the launch period, most traffic during initial spike
    • Write a newsletter. (directive? informational?)
    • have a promotion plan, recruit from outside zooniverse
    • use talk during initial launch (all)
    • moderators for talk (all)
  • in for the long hall
    • reminder emails (all)
    • create community--more commitment from volunteers involved in talk (all, social)
    • positive feedback, report on accomplishments (transactional)

notes from observations (classification shared doc)

  • question level vs. board/discussion level
  • scientist feedback and volunteer feedback--supports shared leadership
  • most feedback isn't exactly "leadership". responding to requests for information, information sharing, not exactly directive leadership.
  • single response vs multi-response questions
  • questions from research team vs questions from public
  • most chat is response to questions
  • not much traditional "leadership" because instructions are pretty clear/explicit. most discussion above & beyond the project
    • only examples of transactional (and possibly directive) leadership
    • we might find examples of social leadership, but likely in smaller projects
  • could "questions asked" be an independent variable?
  • is an informational post directive leadership? is it directive leadership when it responds to a question?
    • information sharing is an important component of engaging volunteers, but isn't necessarily "leadership"
    • can we find lit on information sharing? communities of practice?
  • responses to questions might be a good indication of leadership
    • sometimes responses are just informational
  • types of contributors
    • volunteer
    • moderator
    • scientist
    • zooniverse employee
    • admin (spacewarps)

independent vars

  • levels: question, board, project
  • question lvl vars (possibly omit this level because it's too complicated and not that helpful):
    • is question: 0/1
    • response (is two way communication): 0/1
    • response time
    • poster role
    • answer role
    • leadership? (transactional, aversive, directive, person focused)
  • discussion lvl vars:
    • object/discussion
    • num posts
    • num contributors
    • num responses (in response to)
    • avg response time
    • num volunteers, num mods, num scientists, num zooniverse team
    • num leadership (transactional, aversive, directive, person focused)
  • project level vars
    • num volunteers, num mods, num scientists, num zooniverse team
    • num discussions
    • num objects
    • project length (time)
    • frequency of blog posts
    • length of blog posts