Concepts · Chapter 2

The ideas behind the questions

The study distinguishes what a tool can do from whether it fits the work, and examines the responses and human effort needed to make it usable.

Key concepts and distinctions

Capability

What a model can attempt: drafting, summarising, comparing from a prompt.

Workability

Whether a task can be carried through in a form that is sufficiently usable, understandable and accountable for its setting. It does not mean full automation or that it would work elsewhere.

Ordinary inconvenience

A difficulty that does not change how the task proceeds.

Sociotechnical friction

A poor fit between tools, information, the way work is organised, task requirements and responsibility that changes how the work proceeds.

Local adaptation

Any goal-directed adjustment to a practical constraint: in use, governed, provisional, exploratory, proposed or for one task.

Workaround formation

The narrower case where an alternative path became part of the continuing work needed to complete a task.

Less visible work

Preparation, coordination, checking, correction, explanation, support and responsibility behind an output. It may still be an expected, recognised part of a role.

Invisible labour

The narrower case where necessary time, skill or responsibility was not adequately allocated, supported or recognised.

These are the thesis's working definitions, built from the literature below. Thesis Sections 1.8 and 2.9

Why no single perspective was enough

Perspective	Helps explain	Leaves open
Technology in practice Trist; Orlikowski	How a poor fit becomes consequential for the work	Which response follows
Workarounds and adaptation Alter; Azad and King	How an alternative becomes part of continuing work	What work continues around it
Articulation work and labour Strauss; Bainbridge; Simkute et al.; van den Broek	The preparation, checking and explanation behind outputs	Whether that work is invisible labour

Gasser's account of local fitting runs through all three. Research on preparing information also helps explain why data preparation and traceability mattered in the SDF workflow. Thesis Sections 2.8 and 7.7.1

Literature and how the thesis uses it

The research draws on studies of technology in practice, workarounds and the human work surrounding automated outputs. Select an author to view the full reference.

Why GenAI capability is not enough

Source	What it shows	How the thesis uses it
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D. M., Wu, J., Winter, C., ... Amodei, D. (2020). Language models are few-shot learners [Preprint]. arXiv. https://arxiv.org/abs/2005.14165	Language models can perform varied tasks from instructions and examples in a prompt.	Why GenAI became relevant to workplace tasks.
Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C. L., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., Schulman, J., Hilton, J., Kelton, F., Miller, L., Simens, M., Askell, A., Welinder, P., Christiano, P., Leike, J., & Lowe, R. (2022). Training language models to follow instructions with human feedback [Preprint]. arXiv. https://arxiv.org/abs/2203.02155	Training with human feedback improves how well models follow instructions.	Following instructions well is not the same as a dependable workflow.
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., Brynjolfsson, E., Buch, S., Card, D., Castellon, R., Chatterji, N., Chen, A., Creel, K., Davis, J. Q., Demszky, D., ... Liang, P. (2021). On the opportunities and risks of foundation models. Center for Research on Foundation Models, Stanford University. https://crfm.stanford.edu/report.html	The effects of foundation models depend on data, adaptation, evaluation, deployment and risk.	Having a model available does not settle whether work is workable.
Dell’Acqua, F., McFowland, E., III, Mollick, E., Lifshitz, H., Kellogg, K. C., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2026). Navigating the jagged technological frontier: Field experimental evidence of the effects of artificial intelligence on knowledge worker productivity and quality. Organization Science, 37(2), 403–423. https://doi.org/10.1287/orsc.2025.21838	AI assistance has uneven effects across knowledge-work tasks that look related.	Judge workability task by task.
Feldman, M., Friedler, S. A., Moeller, J., Scheidegger, C., & Venkatasubramanian, S. (2015). Certifying and removing disparate impact. In Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 259–268). Association for Computing Machinery. https://doi.org/10.1145/2783258.2783311	A removed attribute can remain inferable from other data.	Removing identifiers does not make data neutral.
Simkute, A., Tankelevitch, L., Kewenig, V., Scott, A. E., Sellen, A., & Rintel, S. (2025). Ironies of generative AI: Understanding and mitigating productivity loss in human-AI interaction. International Journal of Human-Computer Interaction, 41(5), 2898–2919. https://doi.org/10.1080/10447318.2024.2405782	GenAI can shift work from producing material to evaluating and correcting it.	Review is part of workability, not an optional extra.
van der Maden, W., Sadek, M., Xiao, Z., Mottelson, A., Liao, Q. V., & Zhu, J. (2026). Results-actionability gap: Understanding how practitioners evaluate LLM products in the wild. In Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (Article 18, pp. 1–17). Association for Computing Machinery. https://doi.org/10.1145/3772318.3791069	Evaluating LLM products involves checks, expert review, feedback and support.	An output is not the end of the process.
Zirar, A., Ali, S. I., & Islam, N. (2023). Worker and workplace artificial intelligence (AI) coexistence: Emerging themes and research agenda. Technovation, 124, Article 102747. https://doi.org/10.1016/j.technovation.2023.102747	Workplace AI raises themes of trust, augmentation, skills and human involvement.	Ask who uses, checks and is responsible.

Technology in practice (RQ1)

Source	What it shows	How the thesis uses it
Trist, E. (1981). The evolution of socio-technical systems: A conceptual framework and an action research program (Occasional Paper No. 2). Ontario Ministry of Labour, Ontario Quality of Working Life Centre.	Social and technical arrangements are organised around a primary task.	Judge workability at the level of the whole work system.
Orlikowski, W. J. (1992). The duality of technology: Rethinking the concept of technology in organizations. Organization Science, 3(3), 398–427. https://doi.org/10.1287/orsc.3.3.398	A technology’s role cannot be read from its design or availability.	Access to a tool does not determine its fit.
Orlikowski, W. J. (2008). Using technology and constituting structures: A practice lens for studying technology in organizations. In M. S. Ackerman, C. A. Halverson, T. Erickson, & W. A. Kellogg (Eds.), Resources, co-evolution and artifacts: Theory in CSCW (pp. 255–305). Springer. https://doi.org/10.1007/978-1-84628-901-9_10	Technologies, practices and arrangements are enacted together in use.	Look at what people do and what stays outside the tool.
Hardwicke, N., & Riemer, K. (2018). Do you understand our understanding? Personas as hermeneutic tools in social technology projects. ACIS 2018 Proceedings, 89. https://doi.org/10.5130/acis2018.dn	Software roles can emerge through situated engagement.	Why similar tools can have different statuses.
Veles, N., Graham, C., & Ovaska, C. (2023). University professional staff roles, identities, and spaces of interaction: Systematic review of literature published in 2000–2020. Policy Reviews in Higher Education, 7(2), 127–168. https://doi.org/10.1080/23322969.2023.2193826	University professional work crosses units, roles and relationships with academics.	Why GenAI work crosses boundaries in a university.
Gasser, L. (1986). The integration of computing and routine work. ACM Transactions on Office Information Systems, 4(3), 205–225. https://doi.org/10.1145/214427.214429	Computing becomes workable through local fitting and adjustment.	A thread running through all three questions.
Buck, C., Doctor, E., Eymann, T., & Simoes, E. J. (2020). A systematic literature review on antecedents of workarounds related to information systems in hospitals. In N. Gronau, M. Heine, H. Krasnova, & K. Pousttchi (Eds.), Entwicklungen, Chancen und Herausforderungen der Digitalisierung: Proceedings der 15. Internationalen Tagung Wirtschaftsinformatik (WI 2020) (pp. 664–679). GITO Verlag. https://doi.org/10.30844/wi_2020_f7-buck	Conditions behind workarounds, such as poor fit and inflexibility, occur together.	Reasons for a response, not its status.
Sui, Y., Zou, J., Zhou, M., He, X., Du, L., Han, S., & Zhang, D. (2024). TAP4LLM: Table provider on sampling, augmenting, and packing semi-structured data for large language model reasoning. In Y. Al-Onaizan, M. Bansal, & Y.-N. Chen (Eds.), Findings of the Association for Computational Linguistics: EMNLP 2024 (pp. 10306–10323). Association for Computational Linguistics. https://doi.org/10.18653/v1/2024.findings-emnlp.603	How table-like data is prepared shapes what a model can do with it.	Friction can begin before a question is asked.

Workarounds and adaptation (RQ2)

Source	What it shows	How the thesis uses it
Alter, S. (2014). Theory of workarounds. Communications of the Association for Information Systems, 34(1), 1041–1066. https://doi.org/10.17705/1CAIS.03455	A goal-directed change that lets work continue despite obstacles; temporary or extended.	The basis for workaround formation, with a narrower scope.
Azad, B., & King, N. (2008). Enacting computer workaround practices within a medication dispensing system. European Journal of Information Systems, 17(3), 264–278. https://doi.org/10.1057/ejis.2008.14	Workarounds become usable when fitted to roles, timing and cooperation.	Building something is not the same as it becoming part of the work.
Petrides, L. A., McClelland, S. I., & Nodine, T. R. (2004). Costs and benefits of the workaround: Inventive solution or costly alternative. International Journal of Educational Management, 18(2), 100–108. https://doi.org/10.1108/09513540410522234	Higher-education staff build informal arrangements when systems fall short.	Their status depends on whether they enter continuing work.
de Vargas Pinto, A., Maçada, A. C. G., & Mallmann, G. L. (2018). Workaround behaviour in information systems research. Revista de Gestão, 25(4), 430–446. https://doi.org/10.1108/REGE-12-2017-0015	Workarounds are an organisational issue with wider effects.	Responses can create dependencies and risks.
Behrens, S., & Sedera, W. (2004). Why do shadow systems exist after an ERP implementation? Lessons from a case study. In PACIS 2004 Proceedings (Paper 136). https://aisel.aisnet.org/pacis2004/136	Shadow systems run alongside legitimate systems.	Judge a local system by what it does, not where it came from.
Ciborra, C. (2002). The labyrinths of information: Challenging the wisdom of systems. Oxford University Press.	Improvisation, tinkering and drift are normal in systems in use.	Responses develop through practical engagement.
Wibisono, A., Alhassan, I., Sammon, D., Heavin, C., Kiely, G., & Suryani, E. (2019). Understanding theory of workarounds in practice. Procedia Computer Science, 161, 187–194. https://doi.org/10.1016/j.procs.2019.11.114	Workarounds as a process from obstacle to adjusted work.	Telling provisional changes from incorporated ones.

Continuing human work (RQ3)

Source	What it shows	How the thesis uses it
Strauss, A. (1985). Work and the division of labor. The Sociological Quarterly, 26(1), 1–19. https://doi.org/10.1111/j.1533-8525.1985.tb00212.x	Articulation work keeps tasks, people, records and responsibilities aligned.	Coordination around GenAI-supported outputs.
Bainbridge, L. (1983). Ironies of automation. Automatica, 19(6), 775–779. https://doi.org/10.1016/0005-1098(83)90046-8	Automation can leave people responsible for monitoring and intervention.	Technical change moves human work around.
van den Broek, E. (2025). Unpacking AI at work: Data work, knowledge work, and values work. Information and Organization, 35(3), Article 100584. https://doi.org/10.1016/j.infoandorg.2025.100584	Data work, knowledge work and values work surround AI outputs.	Identifies what work is 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	Data work can disappear behind later outputs.	Preparation behind outputs.
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	Trust in data outputs is built through collaborative work.	The work of building trust is less visible work.

Preparing information for GenAI

Source	What it shows	How the thesis uses it
Fang, X., Xu, W., Tan, F. A., Zhang, J., Hu, Z., Qi, Y., Nickleach, S., Socolinsky, D., Sengamedu, S., & Faloutsos, C. (2024). Large language models (LLMs) on tabular data: Prediction, generation, and understanding – A survey. Transactions on Machine Learning Research. https://openreview.net/forum?id=IZnrCGF9WI	Using language models with tables depends on how tables are prepared and represented.	Preparation is not neutral formatting.
Sculley, D., Holt, G., Golovin, D., Davydov, E., Phillips, T., Ebner, D., Chaudhary, V., Young, M., Crespo, J.-F., & Dennison, D. (2015). Hidden technical debt in machine learning systems. In Proceedings of the 28th International Conference on Neural Information Processing Systems (Vol. 2, pp. 2503–2511). MIT Press. https://proceedings.neurips.cc/paper/2015/hash/86df7dcfd896fcaf2674f757a2463eba-Abstract.html	Hidden dependencies make machine-learning systems hard to maintain.	Why traceability mattered in the SDF workflow.
Zinkevich, M. (2025, August 25). Rules of machine learning: Best practices for ML engineering. Google for Developers. https://developers.google.com/machine-learning/guides/rules-of-ml	Consistency across data, transformations and outputs supports reliability.	Keeping prepared files inspectable.
Liu, N. F., Lin, K., Hewitt, J., Paranjape, A., Bevilacqua, M., Petroni, F., & Liang, P. (2024). Lost in the middle: How language models use long contexts. Transactions of the Association for Computational Linguistics, 12, 157–173. https://doi.org/10.1162/tacl_a_00638	Models use information less reliably in long, unfocused inputs.	Bounded, focused prepared files.
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks [Preprint]. arXiv. https://arxiv.org/abs/2005.11401	Responses can be conditioned on retrieved external sources.	A limited comparison only: no retrieval system was built.

Method sources (Anderson; O’Riordan; O’Raghallaigh et al.; Huber; Wall; Edwards; Sparkes; Braun and Clarke; Fereday and Muir-Cochrane; Nowell et al.) are on the method page; every source is in the reference list.