PaperlayerView exact revision on arXiv

Source-linked AI summary

Workplace Ambiguity and Technology Support for Autistic and IDD Workers

Silas Curry, Rishit Singh, Heekyoung Jung, Angel Hsing-Chi Hwang, Fan Yang, Ce Zhong, Nicholas Lindgren, Chen Chen, Lingyao Li, Renkai Ma

arXiv:2610.21236v1cs.HCcs.CY

TL;DR

Workplace ambiguity leaves autistic and IDD workers acting on unstated expectations, sometimes before needed information becomes legible. The study used caregiver input to explore technology supports and synthesized workplace ambiguity into three intersecting dimensions.

  • Problem

    Autistic and IDD workers must act on unstated workplace expectations, which may become explicit only after they have guessed.

    Source: §1 · pp. 2–3

  • Method

    Caregivers used a large language model in a session to explore and refine technology supports.

    Source: §3.1 · p. 6

  • Results

    Across five themes, workers acted before needed workplace information became legible, and the themes were synthesized into evaluative, procedural, and relational dimensions that feed one another.

    Source: §5 · p. 15, §7 · p. 21

  • Takeaways & Limitations

    The framework treats workplace ambiguity as an access problem rather than an individual inference problem, with employment access renegotiated in everyday interaction.

    Source: §7 · p. 21

    The technology concepts are speculation rather than evidence, and the study heard only one side of the workplace.

    Source: §6 · p. 21

Abstract

from arXiv · show

Workplace ambiguity, where expectations, evaluations, or procedures are left unstated, acts as a structural access barrier for autistic and IDD workers, routinely forcing them to act on guesses. To investigate how these barriers operate, we interviewed 11 caregivers, concluding with an AI co-design activity to explore technological solutions. Their perspectives revealed five themes of workplace ambiguity, which we synthesize into a conceptual framework comprising three intersecting dimensions: evaluative, procedural, and relational. To address these barriers, participants proposed three forms of technology support, including ambient real-time redirection, feedback processing, and physiological guidance, designed to make hidden expectations perceptible and actionable. However, this exposes a core tension: technologies designed to make the workplace legible can inadvertently make the worker hyper-legible to employers through surveillance. We argue that ambiguity-aware workplaces must make expectations explicit before they carry consequences, and assistive technologies must remain accountable to the worker.

1 Introduction

The introduction frames unstated workplace expectations as a post-hiring access barrier for autistic and IDD workers. It presents caregiver interviews and an AI co-design activity as ways to examine the ambiguity and possible technology supports.

1 Introduction

  • Autistic and IDD workers may be judged against expectations that were never stated, leaving them to act before necessary workplace information becomes clear.The introduction frames this as a barrier inside the workplace, after hiring.Source: §1 · p. 2
  • The study asks caregivers what workplace ambiguities they observe and what technology supports they consider useful.Caregivers may see both sides of employer exchanges and notice information or support missing before a worker is judged.Source: §1 · pp. 2–3
  • The authors interviewed 11 caregivers and employment-support professionals, then used a generative-AI co-design activity to explore technology supports.The interviews addressed workplace ambiguity; the co-design activity addressed possible support.Source: §1 · p. 3
  • Caregiver accounts yielded five themes: hidden social norms, unclear performance expectations, misjudged social boundaries, opaque formal structures, and barriers to support.The themes are synthesized into three intersecting dimensions: evaluative, procedural, and relational ambiguity.Source: §1 · p. 3
  • Participants proposed ambient real-time redirection, feedback processing, and physiological guidance as technology supports.These supports shift when information reaches a worker and who is empowered to interpret it.Source: §1 · p. 3
  • Tools that make workplaces legible can also make workers hyper-legible through surveillance and forced disclosure.The authors argue that expectations should be explicit before consequences and that assistive technologies should remain accountable to workers.Source: §1 · p. 3

2 Related Work

Prior work describes workplace barriers as situations where consequential expectations, social norms, performance judgments, or support processes remain unclear. The literature motivates an ambiguity lens and identifies a gap in technology designed for workers navigating ambiguity in employment settings.

2.1 Workplace Barriers Motivating an Ambiguity Lens

  • Autistic and IDD workers may be expected to act while information about expectations, social norms, performance, or support remains unclear.Prior work links unclear expectations and sensory environments to strain for workers and employers, and describes explicit expectations and accommodations as features of good-fit workplaces.Source: §2.1 · p. 4

2.2 Workplace Ambiguity among Autistic and IDD Workers

  • Workplace ambiguity spans four initial dimensions: instructional tasks, social norms in interactions, evaluative feedback, and navigation of supports.The paper carries these literature-based categories into its interview protocol.Source: §2.2 · pp. 4–5
  • Workplace ambiguity includes uncertainty about how performance is judged and about processes for accessing accommodations, rights, and organizational support.Workers may have to identify and request adjustments without recognizing which accommodations they are entitled to.Source: §2.2 · pp. 4–5

2.3 Technology Support for Autistic and IDD Users

  • Technology support for ambiguity in actual employment settings remains poorly understood, despite existing tools and support networks.Prior tools such as planners and visual schedules externalize steps, while caregivers and employment-support professionals participate in navigating workplace barriers.Source: §2.3 · p. 5

3 Methods

The study interviewed 11 caregivers supporting autistic and IDD workers and used an AI-supported co-design activity tied to workplace episodes they had described. Researchers analyzed interview and co-design data using reflexive thematic analysis.

3.1 Data Collection

  • The study interviewed 11 caregivers supporting autistic or IDD people in current or recent paid, volunteer, or work-like training settings.Caregivers included family members and employment-support professionals, and their accounts were not treated as substitutes for workers’ own experiences.Source: §3.1 · pp. 5–6
  • Interviews elicited concrete episodes of unclear expectations, social cues, feedback, or support, then asked caregivers to design technology for an episode they had recounted.Participants used a large language model to explore and refine support ideas, review its responses, and critique or revise its outputs.Source: §3.1 · p. 6

3.2 Data Analysis: Reflexive Thematic Analysis

  • Researchers used reflexive thematic analysis, primarily inductive, while using four initial ambiguity dimensions as a limited deductive lens.Two coders reviewed emerging codes against original quotations during analysis.Source: §3.2 · p. 7
  • Analysis produced five primary themes for workplace ambiguity and three primary themes for technology support.The researchers linked 25 of 54 technology-support codes to more than one ambiguity, then aggregated those links at the theme level.Source: §3.2 · p. 7

4 Findings

The findings are organized around workplace-ambiguity themes and technology-support themes, with a table linking the two. The paper distinguishes analytic dimensions from themes produced by participants’ perspectives.

4 Findings

  • The findings report five themes of workplace ambiguity and three themes of technology support.The technology-support themes are related to the ambiguity themes in the findings.Source: §4 · p. 7
  • Table 2 maps each technology-support theme to the workplace-ambiguity themes it addressed.Its rows are support themes, its columns are ambiguity themes, and its final column reports theme reach and participant–ambiguity links.Source: §4 · p. 7
  • The paper uses dimensions for analytic categories carried into and out of the study, and themes for patterns produced from participants’ perspectives.Source: §4 · p. 7

4.1 RQ1: Themes of Workplace Ambiguity

This section examines how workplace ambiguity and uncertain support shaped autistic and IDD workers’ ability to navigate workplace expectations, structures, and social boundaries. It covers barriers arising from withheld information, gatekeeping, stigma, and insufficient or intrusive support, as well as the need for timely, legitimate, autonomy-preserving support.

4.1 RQ1: Themes of Workplace Ambiguity

  • When expectations were conveyed by inference rather than statement, wrong guesses could bring discipline and asking for clarification carried social costs.Under these conditions, withheld information operated as an access barrier rather than an ordinary workplace annoyance.Source: §4.1 · p. 8

4.1.1 The Hidden Curriculum of Workplace Social Norms.

  • The hidden curriculum left workers to infer informal rules and how others would interpret their behavior, while colleagues still used those norms to evaluate them.One worker had to manage technical and social demands at once, and friction with a secondary workplace relationship contributed to a performance improvement plan despite positive supervisor feedback.Source: §4.1.1 · p. 8
  • Local norms could make literal rule-following a double bind: workers were praised for formal compliance but penalized when local culture expected rules to be softened.Participants contrasted clear break limits with unwritten norms and described safety reports about blocked exits and moldy products becoming socially risky.Source: §4.1.1 · p. 8

4.1.2 Difficulty in Evaluating Performance Expectations.

  • Performance expectations became a moving target when standards, changing instructions, feedback timing, and help-seeking norms were communicated indirectly.Workers could lack procedural anchors such as task sequence and stopping points, hesitate to ask for clarification, and receive feedback too late to connect it to the behavior.Source: §4.1.2 · pp. 8–9
  • Expectations varied across trainers, task shifts, and workplace rhythms, and workers could be judged to have failed before standards became legible.Participants also described unclear expectations interacting with time awareness, overwhelm, or emotional overcorrection, with confusion sometimes interpreted as attitude.Source: §4.1.2 · p. 9

4.1.3 Misjudgment of Social Boundaries with Non-Disabled Peers.

  • Social boundaries were often conveyed through subtle cues, leaving workers to infer when to end or change interactions and how much discretion they had.Participants described missed or overly literal readings of cues, and boundary miscalibration could be handled through discipline rather than mentoring.Source: §4.1.3 · p. 9

4.1.4 Opacity of Formal Workplace Structures.

  • Formal workplace structures were inaccessible when workers had to infer roles, authority, procedures, or how to make requests without being shown the pathways.Roles could be split across people, while accommodation processes assumed procedural literacy: policies existed, but workers were not taught how to request or transfer support.Source: §4.1.4 · pp. 9–10
  • Even when a support need was recognized, access could depend on navigating gatekeepers and uncertainty about whether help was allowed, resented, or sustained as circumstances changed.Source: §4.1.5 · p. 10

4.1.5 Barriers to Accessing Workplace Support.

  • Access depended on HR and supervisors recognizing requests as legitimate, while institutional pushback and blocked advocate contact could exhaust workers and contribute to leaving.Source: §4.1.5 · p. 10
  • Visible support could invite stigma, as coworkers questioned extra help and judged workers by non-disabled standards, resolving one ambiguity while creating another about receiving help without stigma.Source: §4.1.5 · p. 10
  • Support could be too thin or intrusive: fragmented coaching missed ambiguity during shifts, families filled gaps in onboarding, and overcompensation could prevent workers from building confidence.P6 cautioned that people who jump in too quickly may deny workers the opportunity to experience and work through uncertainty.Source: §4.1.5 · pp. 10–11
  • Ambiguity persisted when support was reactive, socially stigmatized, or dependent on individual advocates instead of workplace routines, so participants emphasized timely, legitimate, autonomy-preserving support.Source: §4.1.5 · p. 11

4.2 RQ2: Technology Supports for Workplace Ambiguity

This section examines caregiver-proposed technologies for making workplace expectations and communication more actionable while supporting workers’ self-regulation. It also considers physiological guidance, emphasizing that bodily signals are uncertain and that their use should remain under worker control with safeguards against surveillance.

4.2 RQ2: Technology Supports for Workplace Ambiguity

  • The three support themes were real-time cueing, feedback processing, and socio-emotional guidance; 25 of 54 RQ2 codes linked a support to more than one ambiguity.Participants saw changes to when workplace information arrives and who interprets it as reaching several ambiguities at once.Source: §4.2 · p. 11
  • Real-time cueing and feedback processing each reached all five ambiguity themes, while physiological support reached only performance expectations and social boundaries.Cueing concentrated on social norms, performance expectations, and social boundaries; feedback processing accumulated the most links and was proposed by multiple caregivers in every ambiguity column.Source: §4.2 · p. 11

4.2.1 Ambient Real-Time Behavioral Redirection.

  • Discreet real-time cues could make timing, boundaries, task transitions, and sensory limits perceptible before a supervisor or customer reacted negatively.Participants cautioned that repeated alarms could be disabled and overly literal prompts could lead to overcorrection.Source: §4.2.1 · pp. 11–13

4.2.2 Technology-Mediated Feedback Processing.

  • Feedback-processing tools could translate unclear communication into actionable knowledge and shift design toward improving communication between workers and workplaces.Proposals included feedback translation, performance check-ins, policy and procedure interfaces, and tools to help managers phrase instructions clearly.Source: §4.2.2 · pp. 13–14

4.2.3 Physiologically Informed Socio-Emotional Guidance.

  • Physiologically informed guidance could help workers notice escalation and pause before difficult interactions, while bodily data could help supporters identify recurring stress patterns.A prompt might flag an unusually high heartbeat and suggest a break before speaking with a supervisor; post-workday patterns could identify anxiety spikes around breaks or task transitions.Source: §4.2.3 · pp. 14–15

4.2.3 Physiologically Informed Socio-Emotional Guidance

  • The same physiological signal may indicate anxiety, excitement, effort, or environmental stress depending on the person and moment.Source: §4.2.3 · p. 15
  • Participants called for individualized thresholds and worker control over who can see physiological data to prevent guidance from becoming workplace surveillance.Source: §4.2.3 · p. 15
  • Bodily data may help workers and supporters identify patterns only when systems treat interpretation as uncertain, personal, and accountable to the worker.Source: §4.2.3 · p. 15

5 Discussion

The discussion frames workplace ambiguity as an access problem when unstated expectations and others’ interpretations are not legible before consequential action, across evaluative, procedural, and relational dimensions. It considers technology that can clarify workplace information while creating surveillance and misinterpretation risks, and calls for explicit expectations and support that remains accountable to autistic and IDD workers.

5 Discussion

  • Workers often had to act before needed information became legible, while the three technology themes changed when information reached them and who could help interpret it.Source: §5 · p. 15

5.1 Workplace Ambiguity as a Multi-Dimensional Access Problem

  • The five interview themes refined four initial dimensions into evaluative, procedural, and relational ambiguity, with instructional ambiguity divided across the first two.The dimensions are inseparable because others may interpret workers’ behavior while evaluative or procedural information remains unclear.Source: §5.1 · p. 16
  • Workplace access continues after hiring through interactions that shape workers’ standing and later access to feedback or support, so ambiguity shifts the burden from worker inference toward organizational practice.Source: §5.1 · pp. 16, 18
  • Evaluative ambiguity concerns workers’ uncertainty about their standing and what feedback requires, including feedback that arrives too late or obscures whether it is a warning.In the clearest case, a formal consequence was the first unambiguous signal that performance had been at risk.Source: §5.1 · p. 16
  • Procedural ambiguity concerns which next action an employer will authorize and whose approval counts, leaving support inaccessible when responsibilities and routes are unclear.Caregivers described reconstructing policy after supervisors or human resources staff blocked or delegitimized support requests.Source: §5.1 · p. 17
  • Relational ambiguity concerns how others interpret workers’ behavior, including unstated social expectations that can shape judgments of competence after the formal task is complete.Source: §5.1 · p. 17
  • Relational interpretations can make evaluative and procedural ambiguity consequential, while fear of how questions or visible support will be received can discourage clarification and support-seeking.Source: §5.1 · pp. 17–18
  • The design goal is to make consequential ambiguity recognizable and actionable before harm, not to provide a complete rulebook; this account is secondhand through caregivers’ interpretive labor.Caregivers’ accounts reveal how interpretation may be shifted onto them while leaving open how workers themselves experience the conditions.Source: §5.1 · p. 18

5.2 Technology Support as Mediation Between Autistic and IDD Workers and Workplaces

  • Real-time cues can make hidden task thresholds and social boundaries actionable before consequences, but may disclose disability, prompt overcorrection, or become behavioral monitoring.Participants proposed discreet cues tied to conversation duration or interpersonal distance; this support clustered around evaluative and relational ambiguity.Source: §5.2 · pp. 18–19
  • Feedback processing distributes clarification across workers and others involved, rather than leaving workers to infer expectations alone, and can support clarification across multiple ambiguity links.Participants proposed translators, employer-facing communication guides, and policy interfaces; workers should control what is disclosed and who is present.Source: §5.2 · p. 19
  • Physiological prompts may help workers pause before high-stakes interactions or notice patterns, but arousal signals do not establish their cause or determine intent, emotion, or policy.The discussion calls for interpretive humility: tools can help workers consider explanations or identify someone who can clarify, while preserving uncertainty and worker accountability.Source: §5.2 · p. 19
  • Workers should control what support technologies do and who can access their outputs; employers must not repurpose support for discipline or productivity monitoring.Caregiver input can help assess fit, but should inform rather than replace worker authority.Source: §5.2 · pp. 19–20, §5.3 · p. 20

5.3 Design and Practice Guidelines for Ambiguity-Aware Workplaces

  • Workplaces should make feedback usable while workers can act on it, specify the applicable standard and adjustment, and distinguish routine coaching from formal concern.Feedback should retain its urgency and authority, and AI may explain it but cannot confirm a supervisor’s intent or formal consequence.Source: §5.3 · p. 20
  • Employers should make authorized routes, applicable procedures, decision-makers, and next steps—including what to do after a refusal—visible and usable.Caregivers may help when invited, but employers remain responsible for accurate procedures; shared process representations can reduce caregivers’ informal advocacy labor.Source: §5.3 · p. 20
  • Workplaces should state the social expectations they evaluate before judging workers, communicate boundaries directly, and offer private clarification without enforcing neurotypical conformity.Caregivers can identify recurring ambiguity when invited, but should not define acceptable behavior on the worker’s behalf.Source: §5.3 · p. 20
  • Workers should control when assistance is active, what information is collected, and who can access it; invited supporters may help configure systems but should not receive data by default or override worker choices.Participants valued discreet, individualized support, while their concerns centered on who would see the data and what conclusions would be drawn.Source: §5.3 · p. 20
  • Employers should prohibit support data from being used for productivity scoring, discipline, or emotion classification.This safeguard addresses concerns about who sees the data and what is concluded from it.Source: §5.3 · p. 20
  • AI systems should distinguish workplace source material from generated interpretation, communicate uncertainty when context is incomplete, and leave workers meaningful authority over how they are represented.Technology supports workplace legibility only when interpretation remains accountable to the worker and does not turn workers into objects of continuous interpretation.Source: §5.3 · pp. 20–21
  • Workplaces should make consequential information available before workers bear the consequences of not knowing it, rather than expecting them to infer what organizations leave unsaid.This information includes socially, evaluatively, and procedurally consequential information made accessible in time to act.Source: §5.3 · p. 21
  • Evaluative ambiguity concerns performance standing and feedback consequences, procedural ambiguity concerns action routes and authority, and relational ambiguity concerns how others interpret behavior.Technology can mediate these conditions but can also reproduce inaccessibility through surveillance, conformity demands, or replacing legitimate communication and organizational responsibility with technological inference.Source: §5.3 · p. 21

6 Limitations and Future Work

Because the study relies on caregivers’ accounts, it describes situations workers encountered and their consequences, but not workers’ internal experiences or conclusions. The technology concepts are speculative, and future work should examine workers’ perspectives and the organizational side of workplace ambiguity.

6 Limitations and Future Work

  • Caregiver accounts document situations workers encountered and their consequences, but not how those situations felt or what workers concluded in the moment.Future work can examine the three dimensions directly with autistic and IDD workers.Source: §6 · p. 21
  • Future work can test whether the dimensions’ described interdependence appears the same from autistic and IDD workers’ perspectives.The study notes that proxy reports can diverge from self-reports, with particularly weak agreement on interpersonal relationships and self-determination.Source: §6 · p. 21
  • The technology concepts are speculative: participants developed them in a single AI-supported session, and none has been built or evaluated.What participants imagined was shaped by the model’s responses.Source: §6 · p. 21
  • The study presents one side of the workplace; supervisors, HR staff, and coworkers appear through participants’ accounts, not their own explanations.Studying those actors could help distinguish neglect, deliberate discretion, or how work is divided across offices.Source: §6 · p. 21

7 Conclusion

Interviews with 11 caregivers identify five workplace-ambiguity themes, synthesized into evaluative, procedural, and relational dimensions. The conclusion frames ambiguity as an access problem and warns that technology intended to clarify work can also expose workers to employers.

7 Conclusion

  • The authors interviewed 11 caregivers and identify five ambiguity themes, synthesized into evaluative, procedural, and relational dimensions.They frame these dimensions as an access problem rather than a matter of individual inference.Source: §7 · p. 21
  • The dimensions feed one another, and employment access is renegotiated in everyday interaction rather than settled at hiring.Source: §7 · p. 21
  • Technology that clarifies a workplace can also be turned on the worker.Source: §7 · p. 21
  • When someone expects a worker to guess what is owed, the decision has already been made; naming that decision makes it arguable.Source: §7 · p. 21

A Interview Protocol

The appendix presents a semi-structured interview protocol combining core questions with selectively used probes. Its AI co-design activity asks caregivers to choose an unclear workplace situation, formulate and review a request, and refine it.

A Interview Protocol

  • Interviewers introduced each domain through an open question rather than presenting the four ambiguity labels.Source: A Interview Protocol · p. 25
  • Core questions were asked of every participant, while interviewers selected probes as each conversation required.The protocol covered workplace confusion and support access.Source: Table 3 · p. 25, A Interview Protocol · p. 25
  • Caregivers chose a specific unclear workplace situation and explained why it exemplified ambiguity.Source: A Interview Protocol · p. 26
  • Participants formulated an AI request, reviewed its usefulness and safety, then considered revisions to make the response more useful.Interviewers suggested refinements only when participants were unsure, such as shorter, more concrete explanations or simpler language.Source: A Interview Protocol · p. 26

B Thematic Codebook and Researcher Positionality

The appendix codebooks document five workplace-ambiguity themes and three technology-support themes. They include representative caregiver quotations and examples of proposed support, including physiological guidance and task-transition prompts.

B Thematic Codebook and Researcher Positionality

  • The RQ1 codebook contains five primary themes and 13 sub-themes developed from 153 initial codes.Source: Table 4 · p. 26
  • Ambiguity categories include workplace-role and procedure confusion, barriers to support, peer stigma, institutional pushback, and gaps in structured support.Caregiver examples include uncertainty about who should contact a supervisor and the absence of regular one-on-ones.Source: B Thematic Codebook and Researcher… · p. 27
  • The RQ2 codebook contains three primary technology-support themes and 10 sub-themes, linked to the ambiguity themes participants said they addressed.Source: Table 5 · p. 27
  • Proposed examples include biofeedback-triggered social guidance and confidence prompts at task transitions.One suggested prompt would flag an elevated heartbeat and suggest taking a break before speaking with a supervisor.Source: B Thematic Codebook and Researcher… · p. 28

C Researcher Positionality

The researchers describe varied personal and professional perspectives informing their interpretation of caregiver accounts. They report using regular discussion to test themes and distinguish observed behavior from inferred internal states.

C Researcher Positionality

  • The authors brought experience from autistic and IDD communities, neurodivergence, family caregiving, and mental-health and behavioral-intervention research.Source: C Researcher Positionality · p. 28
  • Caregivers, rather than workers, supplied the accounts on which the researchers’ interpretation rests.Source: C Researcher Positionality · p. 28
  • During analysis, the researchers coded what participants observed and did rather than what they inferred about workers’ internal states.They discussed their standpoints in weekly meetings and used them to test, rather than confirm, emerging themes.Source: C Researcher Positionality · p. 28
Loading 2610.21236v1…