Sequence authority: an unlearning framework

Abstract

Interfaces can benefit from human repair while leaving it out of the record. When a person compensates for a poor fit, the usable output can confirm the arrangement that required compensation. I call this repair loss. The better the repair, the less visible the need for it.

The paper traces repair loss through sequence authority, the practical priority an interface gives one route through precedence, persistence, or recognition. Repair burden makes that priority visible: the work of noticing a mismatch, introducing another route, sustaining it, and translating it into terms the system accepts.

A trace-based interface autoethnography applies these concepts to three prototype audits: single-name labor, AI-assisted writing, and polysomnography. The comparison shows how a completed document, fluent sentence, or successful clinical test can preserve an outcome while omitting the labor and conditions that made it possible. An encounter changes when a person negotiates another route. An institution changes only when the next encounter no longer depends on the same person carrying the same repair again.

Origin

It started with a piece of corporate rhetoric I heard years ago: learn, unlearn, relearn. I do not remember where I heard it first. It was catchy, and my autopilot mind accepted it as intuitive wisdom. To explore or invent, I thought, I must empty the mental space holding previous norms, traditions, and habits. Unlearning must come before relearning.

At the time, I was working in cinema and beginning to teach it. I was struggling with industry practice and unwritten rules about how a film should be made and how cinema should be taught. I had my own ways of doing both. I tried them. Some failed. Some held. Some produced results I could see. Unlearning became a word I used to describe that struggle.

Then the word became available for almost anything. A company or campaign could attach unlearning to its preferred behavior and make the change sound self-evident. I still wanted to question inherited practice, but the word no longer explained what I was doing. Was unlearning a practice I could describe, or just a persuasive label?

The ordinary meaning suggested undoing what had been learned. Software had trained me to expect undo and redo: reverse, store, restore. A mind does not offer the same command. Neither does a culture or an institution. If prior learning is weakened, where does it go? What persists in the body, file, rule, or reward? Who keeps the record, and who gains authority by naming the new practice?

I needed something I could use before I could explain it: a small protocol. Where does a learned order act? What keeps it in place? What can change? What returns under pressure?

That need moved my attention to interfaces. Cinema had taught me that a medium is not neutral. An interface could be a frame, word, person, tool, institutional category, or idea that mediated action. I first imagined it as a middle point between input and output, like a ride platform between driver and passenger. That analogy failed. The platform does more than connect two ends. It sets access, visibility, price, movement, and recognition. The roles blur. The person operating a tool can be directed by it.

AI made the problem harder. A language model is not only a medium between intention and sentence. It can supply the first durable version of the intention. Better prompting changes the encounter, but it does not remove the training corpus, design choices, or institutional rewards attached to fluent output. Three questions followed: what does the interface place first, what must the person repair, and what disappears when the result looks successful?

Research question

How does an interface make one route easier to enter, continue, or recognize than an alternative, and who must do the work when the given order does not fit?

Three subsidiary questions make the inquiry testable:

1. Which route receives precedence, persistence, or recognition?
2. What must a person notice, explain, repeat, risk, or surrender to introduce another route?
3. What remains in the record after the encounter has been repaired?

Sequence authority and repair loss

An interface is a traceable point where a system’s rules, categories, or expected steps become actionable. It may be an input field, identity record, access credential, billing category, preparation sheet, sensor arrangement, scoring rule, or report.

A route is an ordered course of attention and action within an encounter. A route can be available without being equally actionable. It can also be actionable without being equally recordable.

Repair loss occurs when the conditions or labor that repaired an encounter do not survive in its usable record. A successful outcome can then strengthen the appearance that the original route fitted all along.

An interface-shaped default exists when an identifiable arrangement repeatedly gives one route an advantage over alternatives. I call that advantage sequence authority.

Sequence authority has three forms:

Precedence

One route appears, starts, or receives attention before another.

Persistence

One route is carried from one step to the next without renewed explanation or effort.

Recognition

One route produces an entry, result, identity, or outcome that the system accepts as valid.

These forms can occur together, but they need not. Predictive text can have precedence without controlling the submitted sentence. An institutional identity can have recognition even when the production process occurred elsewhere. A clinical accommodation can change one step while the report preserves the protocol’s categories.

The practical difference becomes visible through repair burden. The person seeking another route may have to:

• notice that the given sequence does not fit;
• formulate an alternative before the system has a place for it;
• translate lived knowledge into an accepted term;
• seek permission and repeat the request;
• accept delay, exposure, reduced access, or an uncertain record.

GIVEN ROUTE

  trigger
     |
  offered
     |
  carried
     |
  recognized


OTHER ROUTE
(repair burden)

  notice
     |
  introduce
     |
  justify
     |
  sustain
     |
  maybe recognized

The normative question begins after the ordering has been described. Sequence authority can support safety, continuity, accessibility, and necessary standardization. It becomes a problem for audit when relevant situated knowledge cannot enter without avoidable extra work, when the burden falls on the person with the least authority to change the arrangement, or when the final record receives the benefit of repair but erases its conditions.

Affordance, script, default, classification, habit, and infrastructure already explain parts of this territory.^[1] Sequence authority adds a narrower question about order: which route becomes actionable, continues, or enters the record first, and where does the work of deviation go?

What unlearning means here

Unlearning has no single agreed definition. Organizational research has described it as reducing the influence of prior knowledge or routines, and sometimes as discarding established routines so that other patterns can develop.^[2] Decolonial and pedagogical work uses unlearning at a different scale: to challenge dominant ways of knowing, inherited positions, and institutional forms of validation.^[3]

This framework keeps the idea of reduced influence but changes the unit: not knowledge or routine in general, but a default carried through an identifiable interface.

Unlearning is the practiced weakening of that default’s sequence authority. The familiar route remains available; what changes is its automatic priority and the work required to depart from it.

This may change what comes first, what is carried forward, what counts as valid, or who must explain the exception. A one-time accommodation leaves the default intact for the next encounter.

Method: trace-based interface autoethnography

Interface autoethnography treats the researcher’s own encounters with systems as situated evidence.^[4] The self is neither a transparent source nor the object of unrestricted confession. The self detects pressure. The artifact bounds the claim.

The unit of analysis is an encounter-carrier pair:

• a dated encounter in which a route became easier or harder;
• an identifiable carrier through which the ordering was enacted or preserved.

The carrier may be a prompt box, identity record, submission workflow, billing term, preparation instruction, sensor arrangement, report field, or file format. The paper uses artifact as the audit field for this carrier.

Admission test

An encounter enters the audit only when the record can identify:

1. a specific situation and date or bounded period;
2. at least two possible routes, even if one was difficult to enact;
3. an identifiable artifact or procedural carrier;
4. an action accepted, altered, negotiated, or declined;
5. an outcome or surviving record;
6. the evidentiary status of each claim.

Without a carrier, route contrast, and surviving trace, the encounter remains reflection rather than an interface audit.

Evidence classes

The audit does not treat all recollection as equally verifiable.

Documentary trace

A version history, message, screenshot, form, report, metadata field, or preparation instruction retained from the encounter.

Contemporaneous note

A field note written during the encounter or within twenty-four hours.

Remembered account

A later recollection, labeled as memory rather than transcript.

Interpretation

The analytic claim made by comparing the first three classes.

Contradictions remain in the record. A formal field and a field note can use different operational definitions, remember different events, or disagree. The audit does not resolve the difference by automatically privileging either institutional documentation or personal memory. It asks what each record was designed to preserve.

Felt effort, embodied data, and cognitive load

Felt effort identifies where to look. It does not establish the mechanism by itself. The move from sensation to method requires a dated encounter, identifiable carrier, recorded action, and surviving result.

Work on felt and embodied data representation argues that relations with data are bodily as well as symbolic.^[5] That work concerns the design of representation, not the autoethnographic audit of a clinical record. The connection here is limited but useful: a report can preserve measurements while the body also registers effort, interruption, position, sound, and time. The audit compares those records without treating either as complete.

Cognitive load is another place where sequence authority becomes practical.^[6] A dense report, heavy file, crowded screen, or poorly marked structure can decide who remains in an encounter long enough to act. The burden is uneven for a reader who is tired, interrupted, multilingual, low-vision, older, low-bandwidth, anxious, or working with limited memory. A heading, semantic structure, lighter file, or shorter route does not remove intrinsic complexity. It can stop presentation from becoming the reason an alternative disappears.

Six audit fields

Reflex

What action felt automatic, urgent, necessary, or safer?

Reward

What benefit, relief, protection, or institutional purpose kept that action in place?

Artifact

What identifiable carrier gave the route precedence, persistence, or recognition?

Reorder

What changed in the sequence, carrier, permission, or rule?

Return

Under what pressure did the familiar route regain priority?

Cost

What exposure, loss, unpaid work, appropriation, or new authority resulted from changing and documenting the encounter?

The fields are recursive. Reordering can create a new reward. Documentation can create a new default. The published audit can become an artifact that gives its author authority.

Counter-cases

The method must retain encounters in which:

• the default improved access or safety;
• the person deliberately preferred the familiar route;
• reordering produced no meaningful change;
• the apparent interface effect was better explained by a resource constraint, embodied need, clinical condition, institutional threat, or personal preference.

A framework that records only successful resistance cannot distinguish analysis from self-confirmation.

Position and safety

The inquiry is situated in the experience of a neurodivergent, multilingual, middle-class Bangladeshi man in Cox’s Bazar who works around humanitarian knowledge systems and domestic care labor.

This position affects what can be noticed and negotiated. English fluency and technical literacy helped me identify a clinical billing term. Neurodivergence makes AI assistance both access infrastructure and a possible source of dependence. In the workplace case, the hidden contributor is male, which departs from the gender pattern emphasized in much of the two-person-career literature without removing employment vulnerability or institutional asymmetry.

Evidence safety sets a further limit. Workplace files can carry confidentiality and access obligations. Prompt logs can contain another person’s information. Medical documents can expose identifiers. The method distinguishes among evidence that exists, evidence that can be retained safely, and evidence that can be published. Public claims must contract when safe evidence does.

Prototype audit 1: single-name labor

Single-name labor is recurring, substantive production performed within another person’s formal occupational role while authorization, pay, accountability, and career evidence remain attached to the authorized employee.

The term does not include every form of domestic support. Emotional care may help a worker continue. Household labor may release time. Informal advice may shape an occasional decision. Single-name labor begins when outside work is repeated, substantive, and incorporated into a deliverable submitted to or evaluated by an institution.

The case draws on recurring backend support around another person’s salaried knowledge-work role. The contribution included research, synthesis, editing, quality assurance, and decisions about structure over a sustained period.^[7]

At first, I called it helping. A deadline would arrive, a document would remain unfinished, and I could see how to repair it. The work moved through conversation, private notes, research, and revision into a file that later looked singular. The better I repaired the deadline, the less reason the completed document gave anyone to ask how it had been made.

Papanek’s two-person single career and Philipps’s unpaid market labor are the closest precedents.^[8] Single-name labor narrows their concern toward contemporary systems of authorization and attribution. A login, email account, personnel record, submission history, and curriculum vitae connect one accountable identity to the work. They are designed to identify authorized responsibility, not to reconstruct every contribution. The problem appears when recurring distributed production is necessary to meet the role’s demands but remains unavailable as evidence of the role’s actual capacity requirements.

Audit

Reflex

Absorb professional production as private care, household necessity, or personal usefulness.

Reward

Income continuity, deadline completion, reduced conflict, institutional approval, and the satisfaction of solving a difficult problem.

Artifact

The authorized identity, account permissions, job description, submission workflow, performance record, authorship field, and file metadata.

Reorder

Separate general support from production of a workplace deliverable. Keep accounts and confidential material with authorized staff. Record recurring capacity gaps without transferring protected content. Redirect production needs toward authorized collaboration, realistic workload allocation, or role redesign.

Return

A deadline arrives, the employee is overloaded, and the hidden contributor can repair the situation quickly. Care restores the single-name route.

Cost

The contributor gains a vocabulary for hidden labor by exposing the vulnerability of the person whose employment the labor helped sustain. Publication can transfer authority from the formal worker to the person who names the arrangement.

What the interface does

The principal form of sequence authority here is recognition. The institution can recognize one employee, one chain of accountability, and one career record. Distributed thinking may occur before submission, but the interface converts the completed product into evidence attached to one role.

Singular responsibility may be legally and ethically necessary. The audit asks when recurring external production can be made visible without breaching security or placing the employee at risk.

Possible traces include a private contribution diary, non-confidential version histories, messages, and a skills record that retains the broad period, type of work, recurrence, learning, and disclosure limit. The public claim cannot exceed what those traces can support safely.

Prototype audit 2: the anticipatory sentence

I use large language models to draft, structure, translate, code, and recover from overload. Assistance can widen access and still acquire sequence authority.

The recurring action is simple: I ask the model first.

The answer arrives before my rough sentence has left a trace. Its fluency then becomes the material to revise. The question is not only what the model writes. It is when its sentence enters the process and which version later counts as the source.

Sometimes the loss is grammatical. Sometimes it is conceptual. A Bangla-shaped rhythm disappears. A hesitation that was doing analytic work becomes a confident transition. A local pressure becomes generic institutional language. The generated paragraph can look like the genre before it has located the argument.

Fluent English carried competence and institutional recognition long before language models arrived.^[9] Research on AI-assisted writing reports both redistributed cognitive work and movement toward more homogeneous styles.^[10] Work on genre signaling separates the recognizable surface of a genre from the rhetorical action it is meant to perform.^[11] Version comparison can test these effects at the encounter scale.

Audit

Reflex

Prompt before recording a phrase, question, Bangla sentence, voice note, or rough structure.

Reward

Initiation, fluency, speed, cognitive relief, professional polish, translation, and reduced exposure to corrective judgments about English.

Artifact

The blank prompt box, predictive drafting, autocomplete, one-click rewriting, default tone, and the placement of generated text before an original draft.

Reorder

Use one of two access routes. In intention-first use, leave a small trace before prompting. In access-first use, let the model support initiation, then reconstruct the intended claim, ask for alternatives or counterargument, and compare what entered, disappeared, or hardened between versions.

Return

Fatigue, urgency, shame, or fear of professional judgment makes generated fluency feel safer than an unfinished thought.

Cost

Critical AI practice becomes a professional identity of its own. A method intended to resist platform fluency can become another marketable technique, while the platform remains the condition of production.

What the interface does

The principal form of sequence authority is precedence. Generated prose becomes the first durable sentence and therefore the reference point for later revision. Persistence follows when rewrite functions carry its vocabulary and structure through subsequent versions.

The relevant comparison is not human purity against machine contamination. It is among different sequences of supported writing. A useful trace retains:

• any human formulation made before the prompt;
• the prompt and generated response;
• the submitted or published version;
• accepted, rejected, and substantially modified suggestions;
• a note on changes in claim, detail, rhythm, uncertainty, and confidence.

Tool-first use is not automatically a failure. For some encounters, it is the accessible route. The audit fails if it treats dependence as a moral defect or makes unaided initiation the price of authenticity. The narrower aim is to preserve the writer’s capacity to decide what kind of support enters, when it enters, and what it may determine.

Prototype audit 3: the measurable sleeper

The test name is polysomnography.

At a hospital billing counter in Dhaka, the service could not be located until I supplied that technical term. Once the word entered the database, an informal concern became something the institution could recognize, price, and schedule.

Language was the first interface.

The overnight protocol then asked for a measurable sleeper. Preparation instructions set waking, daytime sleep, and arrival conditions. Sensors recorded brain activity, airflow, respiratory effort, oxygen saturation, movement, position, and snoring. I do not normally sleep at the assigned time, and I use familiar sound to settle into sleep. The initial arrangement prioritized clean audio recording over that practice.

I asked whether a familiar soundscape could play at low volume. The technologist consulted the supervising doctor and received permission. I eventually slept after changing to rain with quiet piano.

Audit

Reflex

Try harder to sleep so that the payment, available night, test, and technologist’s labor produce enough data.

Reward

Interpretable data, diagnosis and titration in one session, efficient use of a scarce room and operator, and a direct route to treatment.

Artifact

The billing term, preparation sheet, appointment slot, intake process, sensors, audio restriction, scoring rules, report tables, and recommendation line.

Reorder

Describe an ordinary condition of sleep, ask to introduce familiar audio, and obtain clinical authorization for a local change.

Return

The broader sequence remains. My body must still produce enough scored sleep, wake at the institutional endpoint, and enter the treatment route defined by the report.

Cost

The negotiated exception can become evidence that the system already accommodates difference. This paper can convert a vulnerable medical encounter and a worker’s responsive judgment into authority for its author.

What the interface does

This case combines recognition and persistence. The billing term makes the concern administratively actionable. The protocol carries the patient through preparation, recording, scoring, reporting, and treatment. The negotiated sound changes one condition but not the larger route.

The report diagnosed moderate obstructive sleep apnea. Positive airway pressure substantially reduced the recorded respiratory events during titration.^[12] The audit accepts the clinical value of those findings. It asks what the portable record preserves.

The report assigns zero minutes to sleep latency while my contemporaneous note records prolonged effort to sleep. Its summary records zero toilet visits while the note records repeated visits. Its diagnostic tables assign zero minutes to REM sleep while another sentence places the lowest oxygen saturation during REM sleep.

These differences do not by themselves prove that the clinical report is wrong. Clinical fields, remembered experience, and field notes can use different units and serve different purposes. Their disagreement identifies a translation boundary. The scored report preserves physiological and treatment information. The field note preserves effort, interruption, and the accommodation that helped the test continue.

The local repair succeeded. The final record makes that repair hard to see.

Repair loss across the cases

The central finding is repair loss. The institution, tool, or protocol receives a usable result after a person performs work that the given route did not anticipate. When the repair succeeds, the result can make the route appear adequate. A deadline is met. The prose is fluent. The test produces interpretable data. The need for repair becomes harder to see because the repair worked.

The cases involve different technologies, stakes, and degrees of reversibility. The comparison does not make them equivalent. It locates the same relation between the conditions of production and the record that survives.

The consequential ordering occurs between stages: contribution to attribution, intention to sentence, sensation to measurement. Auditing only the final screen or document misses the transfer.

The repair may be cognitive, linguistic, administrative, emotional, or bodily. Its benefit enters the system. Its history often does not. The audit therefore asks more than whether an alternative was allowed. It asks whether the arrangement learned from the labor that made the alternative workable.

From local reordering to institutional memory

Reordering changes an encounter when it alters precedence, persistence, or recognition. Examples include drafting before prompting, introducing familiar audio before recording, or moving recurring production into an authorized collaboration.

Institutional memory requires more. A local adjustment becomes durable only when some carrier changes: an instruction, field, permission, template, workflow, resource allocation, or decision rule. Otherwise the next person must discover and negotiate the same exception again.

The distinction can be stated plainly:

LOCAL REORDERING
  This encounter no longer
  follows the same order.

INSTITUTIONAL MEMORY
  The next encounter does not
  require the same repair.

An accommodated patient does not establish a patient-centered protocol. A careful writer does not establish a different AI design. A private contribution log does not establish fair workload or attribution.

Traces can inform redesign, collective inquiry, labor negotiation, or clinical review. They do not substitute for the authority of affected people or for the organizational processes required to change rules and resources.

Conversion and capture

Documentation gives an encounter a portable form. A note can become a page, framework, citation, archive, profile, program, or tool. I call this movement conversion.

Conversion is necessary for comparison. It also makes extraction possible. Hidden labor becomes a concept credited to the person who publishes it. A clinical accommodation becomes evidence of the author’s analytic skill. A method for resisting AI fluency becomes content produced with AI.

Capture begins when the portable form:

• hides the exclusions that made it concise;
• presents a negotiated exception as structural change;
• converts another person’s labor or vulnerability into the author’s authority without adequate consent or limitation;
• turns a practice of refusal into a credential while leaving its conditions intact.

Cost remains a required audit field because reordering is not complete when behavior changes. The inquiry must also examine what authority is created by naming, formatting, publishing, and circulating that change.

This paper is part of the problem it names. Versioning does not solve that problem. It keeps the conversion visible and revisable.

Limits

The cases show how the concepts can be used, not how common the patterns are.

The public paper describes traces but does not yet present a corpus of redacted excerpts. A journal article would need a sampling period, documented coding, safe excerpts, disconfirming cases, and a record of how interpretations changed.

Interface order may not be the best explanation for every friction. Scarcity, illness, hierarchy, family obligation, language politics, and preference can sustain the same route. The ability to negotiate and document is also uneven. The method must not turn the less powerful person into the system’s unpaid diagnostician.

Version 9.1 remains a working paper. The next step is a small, ethically bounded corpus of traces analyzed through the same fields.

Audit card

Use the framework on one bounded encounter.

1. Name the route that entered first.
2. Identify what the interface carried forward automatically.
3. Identify what the system recognized as valid.
4. Record who noticed the mismatch.
5. Record the work required to introduce and sustain another route.
6. Change one carrier, permission, or sequence where safe.
7. Record what returned under pressure.
8. Compare the lived encounter with the record that survived.
9. Ask whether the next person would need to perform the same repair.
10. Record what authority or exposure the audit itself creates.

Suggested citation

Zico, Ishtiaque. 2026. “Sequence Authority: An Unlearning Framework.” Working paper, version 9.1.

Working glossary

Unlearning

The practiced weakening of an interface-shaped default’s sequence authority. The familiar route remains available while its automatic priority is reduced.

Route

An ordered course of attention and action within a bounded encounter.

Interface

A traceable point where a system’s rules, categories, or expected steps become actionable.

Interface-shaped default

A route that an identifiable interface repeatedly makes easier to enter, continue, or recognize than alternatives.

Sequence authority

The practical priority an interface gives a route through precedence, persistence, recognition, or a combination of the three.

Repair burden

The work required to notice a mismatch and make another route actionable, sustainable, or recognizable.

Repair loss

The omission of repair work or enabling conditions from the record that survives a successful encounter.

Interface autoethnography

A critical qualitative method that uses situated interface encounters as field sites. The self detects pressure; the artifact bounds the claim.

Single-name labor

Recurring, substantive production performed within another person’s formal occupational role while authorization, pay, accountability, and career evidence remain attached to the authorized employee.

Measurable sleeper

A patient made clinically legible through preparation rules, sensors, scoring categories, and a report, while the person and the conditions of sleep exceed that record.

Conversion

The movement from lived practice into a portable format such as a report, page, framework, citation, archive, profile, program, or tool.

Capture

The absorption of a critical practice into a form that hides exclusions, overstates change, or transfers authority to the person or institution that formats it.

Marginalia

The authorized identity

I did not first experience single-name labor as a theory of attribution. I saw a person I cared about, a deadline, and a problem I knew how to solve. Calling the work invisible is accurate but incomplete. It was also care, competence, urgency, irritation, and the wish to keep a household stable. Those motives made the arrangement easier to repeat.

The institutional identity is not merely an unfair name field. It is a security and accountability boundary. Informal access is not liberation. The final file can still hide how much capacity the role required. To write about it safely, I remove the institution, documents, and many details. Protection creates its own repair loss. The public case becomes cleaner than the lived arrangement.

The sentence before mine

When I write directly in English inside a language-model interface, the machine can correct the thought before I have fully articulated it. The loss is not merely grammatical. A Bangla-shaped rhythm can disappear, but so can the hesitation through which the argument was still changing. A smooth revision can preserve my topic while replacing my question.

I do not want to romanticize difficulty or pretend that an unaided first sentence is more authentic. Some days AI is how a sentence becomes possible. The practice I want is smaller: leave enough of a trace to know whether the model assisted my thought, redirected it, or supplied a more recognizable thought in its place.

The repaired night

The billing word opened the service. The negotiated sound opened sleep. I remember trying one familiar recording and then changing to rain with quiet piano while the sensors remained attached. I did not refuse measurement. I needed one condition of my ordinary sleep to enter the room so measurement could happen.

The technologist did not simply break a rule. He used discretion to preserve the protocol’s purpose. The report then carried the diagnosis and treatment result beyond that room, which was exactly what it needed to do. It did not carry the human repair that made the night recordable. Both facts matter.

The academic object

I wanted practical rules rather than a word that could bless any promised change. The result is now a named framework with fields, definitions, citations, and a suggested citation for itself. Those disciplines make the argument testable. They also make it easier for the paper, and for me, to gain recognition from the problem it describes.

Version 9.1 was revised through sustained AI-assisted dialogue. I supplied the encounters, terms, objections, and decisions. The tool proposed language and structure that I accepted, changed, or rejected. That process does not sit outside the AI case. It is one of its traces. The cleaner this paper becomes, the more important it is to retain that fact.

References

Acker, J. (1990). Hierarchies, jobs, bodies: A theory of gendered organizations. Gender & Society, 4(2), 139-158.

Agarwal, D., Naaman, M., & Vashistha, A. (2025). AI suggestions homogenize writing toward Western styles and diminish cultural nuances. Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, Article 1117, 1-21. https://doi.org/10.1145/3706598.3713564

Akrich, M. (1992). The de-scription of technical objects. In W. E. Bijker & J. Law (Eds.), Shaping Technology/Building Society: Studies in Sociotechnical Change (pp. 205-224). MIT Press.

Banting, S. (2026). Simulated social action: Assessing generative AI’s attempts at the rhetoric of literary criticism. Discourse and Writing/Rédactologie, 35, 78-100. https://doi.org/10.31468/dwr.1149

Bowker, G. C., & Star, S. L. (1999). Sorting Things Out: Classification and Its Consequences. MIT Press.

Ellis, C., Adams, T. E., & Bochner, A. P. (2011). Autoethnography: An overview. Historical Social Research, 36(4), 273-290.

Fanon, F. (1967). Black Skin, White Masks (C. L. Markmann, Trans.). Grove Press.

Flores, N., & Rosa, J. (2015). Undoing appropriateness: Raciolinguistic ideologies and language diversity in education. Harvard Educational Review, 85(2), 149-171.

Grisold, T., Kaiser, A., & Hafner, J. (2017). Unlearning before creating new knowledge: A cognitive process. Proceedings of the 50th Hawaii International Conference on System Sciences, 4614-4623.

Grisold, T., & Kaiser, A. (2017). Leaving behind what we are not: Applying a systems thinking perspective to present unlearning as an enabler for finding the best version of the self. Journal of Organisational Transformation & Social Change, 14(1), 45-60.

Hislop, D., Bosley, S., Coombs, C. R., & Holland, J. (2014). The process of individual unlearning: A neglected topic in an under-researched field. Management Learning, 45(5), 540-560.

Lee, H. P. H., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, Article 1121, 1-22. https://doi.org/10.1145/3706598.3713778

Norman, D. A. (1988). The Design of Everyday Things. Basic Books.

Omizo, R., & Hart-Davidson, W. (2024). Is genre enough? A theory of genre signaling as generative AI rhetoric. Rhetoric Society Quarterly, 54(3), 272-285. https://doi.org/10.1080/02773945.2024.2343615

Papanek, H. (1973). Men, women, and work: Reflections on the two-person career. American Journal of Sociology, 78(4), 852-872.

Pennycook, A. (1994). The Cultural Politics of English as an International Language. Longman.

Perovich, L. J., Vidrin, I., & Zizzi, N. (2025). Where the embodiment is: Designing felt representations of data. Interactions, 32(5), 38-43. https://doi.org/10.1145/3747171

Philipps, L. (2008). Silent partners: The role of unpaid market labor in families. Feminist Economics, 14(2), 37-57.

Polites, G. L., & Karahanna, E. (2013). The embeddedness of information systems habits in organizational and individual-level routines: Development and disruption. MIS Quarterly, 37(1), 221-246.

Salinas, C. (2020). The pedagogy of detachment and decolonial options: Reflections from a minoritized point of view. Nordic Journal of Comparative and International Education, 4(1), 10-25. https://doi.org/10.7577/njcie.3563

Sourati, Z., Ziabari, A. S., & Dehghani, M. (2026). The homogenizing effect of large language models on human expression and thought. Trends in Cognitive Sciences. https://doi.org/10.1016/j.tics.2026.01.003

Star, S. L., & Strauss, A. (1999). Layers of silence, arenas of voice: The ecology of visible and invisible work. Computer Supported Cooperative Work, 8, 9-30.

Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257-285.

Thaler, R. H., & Sunstein, C. R. (2008). Nudge: Improving Decisions About Health, Wealth, and Happiness. Yale University Press.

Tsang, E. W. K., & Zahra, S. A. (2008). Organizational unlearning. Human Relations, 61(10), 1435-1462.

Tufte, T. (2024). Unlearning communication for social change: A pedagogical proposition. Social Sciences, 13(7), Article 335. https://doi.org/10.3390/socsci13070335

Endnotes