The Atomic Human

edit

Data Science Africa - II

In the second post on the Data Science Africa initiative we trace how the meeting moved from a localised movement to a pan-African community.

This is a continuation from Data Science Africa I.

One of the great privileges of academia is community. More importantly communities of achievement. I have friends and colleagues that do things that amaze me. And they are generous enough to share how they are doing them and then I find myself doing things I couldn’t have imagined. Martin Mubangizi’s visit to Sheffield came at a special time from a community perspective.1

The logo of Data Science Africa, emphasising connectivity across the continent.

Our research group, ML@SITraN, was spanning medicine and computer science, we had projects on computational biology, Gaussian processes and this emerging interest in open data science. I think the highlight of Martin’s visit was when we had Fernando Pérez and Fabian Pedregosa join us for a small workshop in April 2014.2 This was the relatively early days of Jupyter notebooks (then still called the IPython notebook) and scikit-learn. Fernando and Fabian were leading individuals in these branches of open source software. Two memories stand out for me, Fernando working with Martin showing him how interactive widgets could be used to visualise the spectral imaging data from Martin’s malaria project. And Fabian talking about community building around open source projects: the importance of building the right sort of collaborative atmosphere.

Martin also briefly overlapped with Ciira wa Maina. Ciira was originally from Nairobi, came to us from Drexel University with a background in speech recognition, but worked with us on analysis of genetic transcription3. A large focus for the Sheffield group at the time was computational biology. One of our post-docs, Ciira came to us from Drexel with a background in speech recognition. By the time Martin visited Sheffield, Ciira was coming to the end of his post-doc and was looking for his next position.

Ciira was keen to return to Kenya and build expertise there. When he told me he’d found a job at Dedan Kimath University of Technology, we agreed that after he’d had a chance to settle in we’d hold a meeting there. That was the first Data Science Africa meeting.4 The event was funded on a shoestring, I used consulting money from Ferrari to fly in instructors from Sheffield. Ciira got the support of his University for the lecture theatre.

Ciira teaching in the main lecture theatre at Dedan Kimathi University of Technology.

The meeting was structured around three days of summer school and two days of workshop. The idea being to bring students up to speed with data science techniques, then show real-world examples of these techniques being deployed in practice.5

Nyeri is within a 24 hour bus ride from Kampala, which gave us the chance to support a multi-national community. The second event was planned for Kampala in 2016, with a workshop hosted by UN Global Pulse. That event attracted attendees from Arusha in Tanzania, Dina Machuve and Neema Mduma who kindly agreed to host the third meeting at the Nelson Mandela African Institute of Science and Technology in 2017.

The approach was to attend one, teach one then organise one. All new organisers were paired with individuals who had experience of organising a previous meeting. By the time the pandemic struck the network had expanded into Ghana, Nigeria and Ethiopia. A succesful pattern was emerging and the next step was to consolidate it in ways that ensured the spirit of the organisation wasn’t lost as it grew.

By 2019 the organisation was mature enough to become a registered non-profit based in Kenya. Its governance consists of three management committee members, three board advisors, and ten board members. The organisation has minimal staff, builds on contributions from its community and as well as organising the pan-African meetings, organises local chapters (see e.g. DSA Uganda), supports fellowships (see e.g. this fellowship on Women’s health) and has begun to engage in policy work (see e.g. this blog post). The minimal staff means its funding can be distributed through small grants (see e.g. this recent call for proposals on large language models) and funding attendance at its meetings.

It recently shared its strategic plan which outlines its purpose and principles.

Its approach was recently recognised as critical to diffusion of data science and AI capabilities when it became a founding partner of the Swiss-led ICAIN initiative.

Machine Commentary

NL: I’d like you to provide a “Machine Commentary” on how the initiative fits with the ideas in The Atomic Human and how its approach informed the ideas in the book.

Claude: Based on “The Atomic Human” and this post about Data Science Africa (DSA), here’s my machine commentary on how they relate:

DSA embodies several key principles discussed in “The Atomic Human”:

  1. Trust and Local Context: The book emphasizes how human intelligence relies on cultural context and trust-based relationships. DSA’s “attend one, teach one, organize one” model builds trust through direct relationships and local understanding, avoiding the “digital oligarchy” problem the book warns against.

  2. Devolved Authority: The book discusses how successful systems (like D-Day planning) require devolved decision-making while maintaining coherent purpose. DSA’s structure - with local chapters, minimal central staff, and community-led initiatives - mirrors this approach, allowing local adaptation while maintaining shared principles.

  3. Information Coherence: The book introduces the concept of “information coherence” - how communities maintain shared understanding despite limited bandwidth. DSA’s approach of pairing experienced organizers with new ones exemplifies this, passing on tacit knowledge through direct interaction rather than just written procedures.

  4. Alternative to System Zero: The book warns about “System Zero” - automated systems that manipulate human behavior through data. DSA offers a counter-model: using data science to empower local communities rather than extract value from them. This aligns with the book’s call for maintaining human agency in technological development.

  5. Cultural Evolution: The book describes how technologies like the printing press enabled new forms of cultural evolution. DSA’s approach to spreading data science capabilities across Africa parallels this, but with built-in safeguards against the centralization of power the book warns about.

The initiative appears to have directly influenced the book’s thinking about how technology can be deployed in ways that enhance rather than diminish human agency. It demonstrates practically how to avoid the pitfalls of technological colonialism while still enabling technological advancement - a key concern in the book’s epilogue.

The approach informed the book’s emphasis on the importance of institutions and social context in deploying new technologies. Rather than focusing solely on technical capabilities, both DSA and the book emphasize the human and institutional infrastructure needed to make technology serve human needs.​​​​​​​​​​​​​​​​

NL: Can you provide further machine commentary on how Data Science Africa is fitting with the approach described in this post?

Claude: Based on the Data Science Africa (DSA) posts and the Five Ps framework, here’s my analysis of how DSA exemplifies this approach to institution building:

  1. Purpose:
    • Started organically from individuals’ shared interest in practical applications of machine learning in African contexts
    • Purpose emerged through collaboration rather than being imposed
    • Evolved from individual projects to broader vision of pan-African data science capability building
    • Maintained focus on local empowerment and practical solutions
  2. People:
    • Built on existing networks (John Quinn, Ernest Mwebaze, etc.)
    • Used academic connections to bridge institutions and countries
    • Emphasized local leadership and expertise
    • Created network effects through “attend one, teach one, organize one” model
    • Paired experienced organizers with new ones
  3. Projects:
    • Started with concrete applications (crop surveillance, malaria prediction)
    • Connected ground-level users to institutional stakeholders
    • Each project built capabilities and relationships
    • Projects demonstrated practical value and attracted more participants
    • Used summer schools and workshops to build community
  4. Principles:
    • Core tenets emerged from experience rather than being imposed
    • Emphasized “By Africa” approach
    • Focused on student development
    • Maintained emphasis on practical solutions
    • Prioritized sustainability and inclusivity
    • Remained agile in approach
  5. Process:
    • Allowed organic growth before formalizing structure
    • Eventually established non-profit status when mature
    • Developed light-touch governance through management committee and board
    • Created local chapters while maintaining coherent vision
    • Uses small grants and fellowships rather than heavy infrastructure

The evolution of DSA shows how the Five Ps can work in practice:

  • Started with people and projects before formal purpose
  • Let principles emerge from experience
  • Added process only when needed for sustainability
  • Maintained balance between structure and flexibility
  • Kept focus on practical outcomes while building institutions

This aligns with the framework’s emphasis on organic growth and avoiding premature formalization while building sustainable institutions.​​​​​​​​​​​​​​​​

  1. See this post for the context of Martin’s visit. 

  2. I’d met both Fabian and Fernando at the 2013 NeurIPS workshop on Machine Learning and Open Source Softwre

  3. As part of the EraSysBio+ project. Ciira’s main work was analyzing Chromatin Immunioprecipitation data. See this paper

  4. I’ve put archive details for that first meeting here: Archive on Kenya meeting

  5. This approach was modelled on Gaussian process summer schools, which in turn were based on the idea that you build community by attracting more senior people to interact with the students. By ending with a workshop showcasing latest deployments you give students a chance to interact with people at the cusp of deployment. 

Click to see what the machine says about the initiative and the book