Reflective journal My account of the SDF work, written in February 2026 about events in 2025.
Technical materials The code, processing logic and diagrams of what I built, showing how it worked.
Dated messages Messages from the time, used only to confirm dates and the order of events.
Participant account What an interviewee described. Not observed or checked independently.
Interpretation dashed outline Conclusions drawn from the evidence.
Stage of use Whether a change was in use, being tested, explored or proposed. These labels describe a stage, not whether the change succeeded.
Research question 3 · Continuing work
Short answer
In all three settings, outputs depended on preparing information, checking, correcting, explaining, learning and coordinating that the finished output did not show. That work was less visible. The evidence did not establish that it was invisible labour, which would also need evidence that it was not given enough time, support or recognition. The reported effects were mixed: some participants described extra review, others described time savings.
Thesis Section 7.5 · full question wording in Section 1.4
Less visible or invisible?
Less visible work
Preparing, coordinating, checking, correcting, explaining, supporting and staying responsible, behind or around an output. It may still be expected, allocated and recognised as part of someone's role.
Further evidence needed
Invisible labour
Necessary time, skill or responsibility that was not adequately allocated, supported or recognised.
Adapted from Thesis Figure 2.6.1 and Section 2.9.3. The same activity can be less visible work in one setting and invisible labour in another, depending on that evidence.
What work remained
The work behind the outputs was not one common task. What people checked depended on the work and on who would rely on the result.
My SDF workflow
Does this figure mean what it seems, given which fund paid for what?
Technical development, preparing information, checking, reconciling and explaining figures to staff.
Reflective journalInstitutional support
Is the data complete, is the application ready, is the draft suitable?
IP01's ongoing warehouse reconciliation (established data work, not shown to be introduced by GenAI); IP02's checks before wider use; IP03's review of drafts in their own mailbox.
Participant accountEveryday staff
Is this generated material right for this task and this person?
Reviewing drafts, correcting information already used, keeping feedback specific to each student, supplying an interpretation the output missed, and learning to use the tools.
Participant account“my work has changed from... spending time thoughtfully writing to now deep review work, which takes a very different critical eye” SP01, learning design
“And then I have to make sure it's accurate. So it's almost a reverse of how I used to work” SP04, teaching and academic work
“Because I got it late, I wasn't in the pilot program, so I kind of trained myself on my own” SP07, strategy-facing work
Thesis Sections 6.7–6.8
These are participants' reports. The study did not measure the overall effect on time or workload. Thesis Sections 6.6–6.7
Invisible labour?
| Setting | Less visible work | Invisible labour | Why |
|---|---|---|---|
| My SDF workflow | Identified | Not established | My records do not show enough about role expectations, workload allocation, support or recognition for this work. Section 4.7 |
| Institutional support | Identified | Not established | IP03 described preparing recipient lists as “invisible time”. The interview did not establish whether this work lacked allocated time, support or recognition. Sections 5.7–5.8 |
| Everyday staff | Identified | No general finding | SP07 reported no set time for training, a reported time-allocation difficulty. SP06's estimate of unpaid work concerned workplace technology generally, and SP08's unpaid peer review was not caused by GenAI. Sections 6.7–6.9 |
Research it builds on
Articulation work: the effort of keeping tasks, people, records and responsibilities aligned.
Automation can leave people responsible for monitoring and stepping in.
GenAI can shift work from producing material to evaluating and correcting it. The cases here add what that checking was about.
Data work, knowledge work and values work around AI outputs. Used to identify what work was present, not its labour status.
Sambasivan, N., Kapania, S., Highfill, H., Akrong, D., Paritosh, P., & Aroyo, L. M. (2021). “Everyone wants to do the model work, not the data work”: Data cascades in high-stakes AI. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems (Article 39, pp. 1–15). Association for Computing Machinery. https://doi.org/10.1145/3411764.3445518 · Passi, S., & Jackson, S. J. (2018). Trust in data science: Collaboration, translation, and accountability in corporate data science projects. Proceedings of the ACM on Human-Computer Interaction, 2(CSCW), Article 136, 1–28. https://doi.org/10.1145/3274405
Data work and the work of building trust can disappear behind later outputs.
Why it matters
Work can be difficult to see while still being recognised as part of someone's role. Understanding whether it is invisible labour also requires knowing how time, support and recognition are provided: who prepares, checks, explains and supports, and whether that work is accounted for. Thesis Sections 7.5 and 7.7.1