Collaborations Workshop 2024

[chartgerink]

I'll be joining the CW24 workshop virtually as part of my dayjob at data.org. This is a place for me to document and if anybody's going as well, feel free to chime in 😊

Schedule.

Purpose: identifying different types of (re)productive work of sustainability in open source.

[chartgerink]

Using the lens of (re)productive work in open source to view the talks at CW24 has been interesting for the first day. I'm sharing some notes below - these are both documenting notes from the talk itself and questions/thoughts of my own.

Arfon Smith - AI/ML tools for science

• Generative AI/LLMs are changing how code is written - in 5 years, the Research Software Engineer role may very well look different as a result
• There was plenty of scepticism in the CW24 room regarding LLMs, few optimistic.
• LLMs are likely to take up a bigger role in
• A study by Accenture on using GitHub Copilot indicated developers stayed in flow state longer using LLMs

• LLMs are tools that can be helpful, but also are "hallucinating tools" - we always need to check what they produce. They may hallucinate helpfully every now and then - according to Arfon the value lies in those moments
• AI will be an interface to more code - it lowers the barriers to entry (i.e., one needs to understand what code to request, not necessarily how to write it yourself)
• Will LLMs do more and more of the reproductive work of software engineering? What does it mean if it explains, teaches, and even produces code for us?
• What does it mean when we can get individualized learning through these hallucination machines (e.g., Explain like i'm 5)? This reproductive work potentially becomes even more individuated than recommender systems
• retrieval augmented generation is a type of AI model that provides context no that ingests context
• Friends dont friends built LLM's

Lightning talks

• Rowland Mosbergen
  • We don't put enough focus on the value/importance of leadership skills (reproductive labor?)
  • Mentions results from the RSE survey
• Sarah Jaffa - Requirements engineering

Lisa Murphy - Open source sustainability

• Seeking funding is a form of reproductive work
• If we set up funding in a way that makes it hard to get funded, it creates more reproductive work
• Not all reproductive work is necessary, but becomes necessary by systemic issues.

Lucy Robinson - Citizen Science

• Citizen science has been going on for hundreds of years, even if it wasn't named as such
  • example includes the documenting of the first cherry blossoms in Japan
• In their work, they renamed citizen science to community science, because it fits better. But renaming doesn't change structural issues.
  • One reflection I have is whether we should call it collective science? Community is being overused in so many domains.
• levels of practice participation in community science
  1. Crowdsourcing
  2. contributory
  3. collaborative
  4. codesigned
• Thought: how does reproductive work show itself in software? How do gender and software interact, taking into account the dimension of productive and reproductive work?
  • The cis male domination may be an origin of the focus on productivity, and undervaluing the reproductive work building relations, communicating, etc.
  • A possible juxtaposition between the professional and the home domain: In the professional domain reproductive work becomes more prevalent as one becomes more powerful/senior, which is joined by forcing those below to be more productive. At home, more power (e.g., in a traditional patriarchal household) is aligned with forcing more reproductive work on the others.
• Lucy indicates three buckets of learning and the keynote
  1. gaining content knowledge
  2. recognising own role in a project/recognize you have expertise to give/recognising that you can contribute
  3. a personal internal sense of agency that you can act
• community science should be framed as do-ers not helper

[chartgerink]

Day 2 running notes are a bit shorter, because there was more discussion work.

• Lots of discussion around the ecological impact of research software engineering and high performance compute
• Mostly around carbon intensity and not on other areas (e.g., water usage, rare earth materials) - the discussion had a fairly individuated sense of ecological impact which can (re)produce feelings of guilt
• At the Turing Way celebration there was talk about using testing as a way to ensure quality:
  • Smoke Testing
  • Unit Testing
  • Integration Testing
  • System Testing
  • Acceptance and Regression Testing
  • Runtime Testing
• Testing is a form of reproductive labor which is abstracted into automated checks - great!
• Underlying some of the discussions was around mental health, and it led me to question whether the excessive focus on productive work is hampering people's mental health, because it also removes time from reproductive work that is more foundational? Maybe I'm seeing nails everywhere 🤷‍♀
• Also fun to merge PRs together, great way to celebrate major work milestones.

Keynote Kelly Widdicks

• Climate impacts of the ICT Sector
• Types
  • Embodied
  • Use phase
  • End of life
• 1.8-2.8% 2.1-3.9% (comparable to aviation)
• Jevon's paradox (aka rebound effect) - efficiency improvements are outpaced by demand
• Being more efficient isn't always effective - what does this mean for everyday work?

• likely that ict is creating rebound effects where it produces emissions
  • e.g., video conferencing has grown alongside flights, so it is additive instead of reductive

• Sustainability in Digital Research Infrastructure: UKRI Net Zero DRI Scoping Project final technical report

[chartgerink]

It is in general incredibly interesting to note that the more I think about it, the more work in software is reproductive, not productive. Lots of work is about recreating conditions of understanding, clarity, and improving relations.

Reproductive	Productive
Scaffolding
Storytelling
Training
Planning
Learning
Recordkeeping

I'm still not sure what all of this means, but given that the event is over I'll leave it here for now.