Grassroots AI - Getting Started Checklist

This is a growing checklist on how to start a distributed collaborative grassroots AI community. It is by no means complete (in fact, this is just a first pass) and is forever changing as different groups include their learnings. Each movement would need to adapt to their individual CONTEXT. That said, we hope some people find value from this

Rationale

There is a growing set of initiatives that want to enable AI research in a participatory and inclusive fashion. To name a few Masakhane, ML Collective, Sisonke Biotik, Delta Analytics, Turkic Interlingua and BigScience.

The activities recommended in the check list are grounded in the following values:

Collaboration
Inclusivity
Trust
Kindness
Scalability
Robustness
Sustainability
Multidisciplinarity

They've largely been evolved in low-resourced environments for volunteer base activities.

Note: Currently the checklist is skewed to the principles used within the Masakhane and Sisonke Biotik initiatives. This means it's likely skewed to the African context

The Checklist

Direction and Momentum

Articulate a long term vision. This is a simple and accessible articulation of why you're organising together.
Initial Collaborative Task. This task should be accessible to beginners and newcomers with a variety of ways to
Exciting Initial Announcement Event. This event serves to spread the word and recruit participants. It is recommended to co-locate this event in a space where their might be a higher density of potential participants, for example: at a related conference. This initial excitement will start the momentum high.
Scheduled Regular Catch up Meetings. These meetings build community, keep momentum going, allow opportunities for feedback, and facilitate partnerships and collaboration.
Milestones - Maintain a public and collaborative list of milestones or goals. This could be a list of conference papers that the movement wants to publish, events the movement wants to speak at.

Collaboration Mediums

Synchronous Communication. A chat platform should be chosen that is familiar to the multidisciplinary stakeholders. This could be anything from Whatsapp groups to Slack to Discord to Teams.
Asynchronous Communication. Not everyone has capacity to keep up with ongoing chats, so an asynchronous communition medium, such as a google mailing list, is excellent for important announcements and discussions.
Meeting Platform. The meeting platform is exellent for regular catch up calls
Open and Accessible Meeting Notes. No meeting time is accessible for all timezones or availabilities. Thus, open and accessible meeting notes are ideal for people to follow.
Open Calendar. All events related to the movement should be included on an open calendar. People can easily subscribe to the calendar and be informed about all events

Inclusivity

Inclusivity requires lowering barriers to entry, but also helping guide newcomers, and grow each others expertise

Simple and Welcoming Onboarding. Make joining the movement and getting access to initial points to start as simple as possible. Avoid requesting lots of details from your participants. e.g. Requesting level of education might make an individual without a traditional scientific education feel less welcome. Provide space for people to introduce themselves and welcome them.
Orientation for Newcomers. Provide an overview of where the movement is at the beginning on each meeting and give anchor points for new participants to start partipating. Having an initial message with some tips to get started also helps significantly.
Platform Familiarity. Consider the accessibility and familiarity of chosen platforms. e.g. Not all disciplines use the same tools, and to achieve multidisciplinarity one may need to choose something less suitable that is more accessible. e.g. Meeting notes on Github (a platform for developers) versus a simple Google Doc (which more people can use).
Code of Conduct. Have a code of conduct in place from day one. The code of conduct should enforce the movement's values.
Peer and Group Mentorship Culture. Peer mentorship acknowledges that each individual has something they can teach another, and Group mentorship provides space for anyone to ask questions and for the members to help each other. Having channel dedicated to people getting help and answering questions is helpful.
Opportunities for Idea Sharing and Feedback. When participants have proposals or ideas, create spaces for them to be shared widely for feedback from the community.
Open Science. Publish preprints, open source code, documentation and datasets, fail and succeed in the open so others can learn
Special Invites to under-represented groups. If you find your group is missing certain participants, then reach out to them individually and invite them to come present their work or arrange a workshop with them.

Robustness, Flexibility, Sustainability and Scalability

Trust and Ownership. When people have an awesome idea on how to make the movement better, provide them with the tools and access so that they can own the initiative. Additionally, trusting others creates more leaders within the movement.
Redundancy in Leadership. Given these sort of movements are typically part-time, they need to be robust to changes in availability in leadership. Growing and enabling multiple community leaders that are given the power to unblock others will allow for the variation in availability. e.g. Empowering and enabling multiple community members to run the regular catchup meetings, so that it's not reliant on a single person's availability, or giving website update access to many community members.
Consistent Engagements. "Do Not Wait To Strike Till the Iron Is Hot; But Make It Hot By Striking". Slow, consistent efforts build. It may seem like nothing is happening, but it is and the consistency matters more. Continuing to meet, despite the odds, is what makes these movements slowly gain momentum.
Embrace Emergent Practices. While top-down structure can be valuable, there aren't any best practices in this space (perhaps only good practices), and so providing individuals with freedom to experiment with their own ideas on how to run the movement or parts of the movement might allow for good practices to emerge.
Dynamic Role Assignment. When people join they come with their own unique set of skills and goals. This means people might join as a linguist but want to become more computational. Allow roles to be dynamic and complex, rather than fixed.

Community

Kinship. A community is greater than the capabilities of the individual. It's the joy of sharing, of celebrating each other and each other's culture. Not all interactions need be about work - create spaces for friendships to build and to get to know each other outside of any initiative.
Celebration and Support. Spur a culture of celebration of each other's wins. This means highlighting successes of the community members (even if not related to the community. e.g. a community member getting a new job). Additionally, provide support to community members who may be having a tough time.
Acknowledgement. Acknowledge people's participation. A community members list on the website is highly effective.
Redefine Authorship. Scientific authorship has traditionally been defined by the Havard Medical Journal. However, this is prohibitively small. In a collaborative movement, each person's contribution is acknowledged: whether they wrote code, annotated data, did community development, provided analysis or guided the writing journey. The paper would not have been possible without each person's contribution, so we recommend to include them as authors.
Organize Large Impactful Events. Arrange workshops, seminars, hackathons for your cause. It brings in new members, revitalizes existing members, and facilitates collaborations

grassroot-ai/getting-started-checklist