Method diffusion in large open-source projects

I'm an alien, I'm a computer scientist in the world of sociology

Today's menu

• Research overview
• Adoptions in Debian
• Future directions

Definitions

Method diffusion in large open-source projects

method

technique, tool, procedure

diffusion

introducing with the goal to cause adoption

large project

possessing a high inertia and complex self-dynamics

Research basis

• Debian's methods evolved over time
• high interdependence of methods
• stagnating efficiency in day-to-day activity
• improved methods exist, but volunteers cannot be told to use them
• under what circumstances would volunteers adopt new methods?

Hypothesis

There exist characteristics of innovations and strategic considerations of their diffusions which directly affect the rate of adoption in volunteer projects

Knowledge of these characteristics enables engineered diffusions to improve efficiency of cooperation

Research approach

two-phased approach:

• identification of attributes of successful diffusions, formalisation of guidelines (case studies, diffusion research)
• application & verification of guidelines (participatory action research)

Phase 1: guideline definition

• case studies
• analyse previous diffusions
• interviews (experience & conceptual)
• usage statistics (from repositories, mailing lists, installation reports)
• (survey data)

Phase 1

The "guidelines" highlight those traits of a diffusion, which have the highest impact on adoption rate.

Phase 2: implementation

• participatory action research
• explain successful/failing diffusions
• predict success of ongoing/planned diffusions
• engineer diffusions for success

Phase 2

Conceptual framework

• skeletal structuring of attributes to facilitate comparisons of diffusions

• existing frameworks from sociology and management science:

  • voluntary, independent adoption decisions by individuals
  • authoritarian, concerted diffusions

• OSS more complex: strong network externalities, ideology, voluntary/cabal

Framework example: Rogers' diffusions of innovations

innovation

relative advantage, compatibility, complexity, verifiability, visibility

communication

hard vs. soft information, mass vs. interpersonal medium, homophily

social system

structure, communication arrangement, social norms, opinion leaders, types of decisions

consequences

desirability, direct vs. indirect, anticipated vs. unexpected

Problems with classical diffusion frameworks

• prior information not required
• low degree of network externalities

thus have to design my own framework, cherry picking from others

very time-consuming

Ideas for OSS-specific factors

• method of development, language chosen
• length of release cycles
• flexibility, accessible for reinnovators
• documentation
• forum activity
• age of bugs and responsiveness of author
• manifesto
• opinion leaders
• multiple interfaces

The adoption process

According to Rogers:

• knowledge
• persuasion
• decision
• implementation
• confirmation

S-shaped curve of adoption speed, success defined by steepness of gradient

Adoptions in Debian

http://wiki.debian.org/madduck/adoptions

Guidelines

• not entirely sure what they will be yet
• ideal would be a list of factors, sorted by degree of impact
  • requires large amounts of data
  • non-qualitative and not really feasible
• think of them as a HOWTO

Engineering for success

• not searching a method to engineer diffusions of just any tool
• method has to solve problems
• assuming a tool maps a sensible workflow, the guidelines will help optimise its diffusion (which would happen by itself)

Implications on management

• treat employees as volunteers
• motivation from (perceived) autonomy

Thank you

Thank you for your attention!

Martin F. Krafft <phd@martin-krafft.net>

• http://martin-krafft.net/phd/
• http://people.debian.org/~madduck/

Licence

These slides, their design, and the content are © Martin F. Krafft and released under the terms of the CC by-nc-sa 2.5 licence.

reStructuredText sources: slides.rst and ui/debian/*.

Slides based on the S5 presentation system.