From Future of Work to Future of Workers: Addressing Asymptomatic AI Harms to Foster Dignified Human-AI Interaction
This is a strong CHI paper because it reframes AI-at-work from efficiency to dignity and expertise preservation, and it does so with a concrete longitudinal account rather than a purely speculative critique. The novelty is less a new algorithm than a new problem framing plus a response framework grounded in observed workplace erosion.
Axes Lens
Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.
Contribution shape
- Knowledge form
- normative knowledge typical · 31/268
- Novelty type
- framework typical · 59/268
- Abstraction level
- practice typical · 85/268
- Generalization target
- organizational context typical · 20/268
- Validation mode
- qualitative study typical · 63/268
Evidence profile
- Evidence strength
- strong typical · 158/268
- Claim alignment
- strong typical · 231/268
- Overclaim risk
- medium typical · 210/268
Review Summary
This paper stands out by shifting the unit of concern from organizational productivity to worker dignity and the preservation of expertise. The central move is persuasive: AI can produce visible operational gains while simultaneously creating hidden, cumulative harms that are easy to miss if one only tracks output metrics. The quoted evidence supports that claim directly through the notion of "intuition rust" and the progression from asymptomatic effects to chronic harms such as skill atrophy and identity commoditization. That is a meaningful departure from common-sense AI adoption narratives, which usually assume that if performance improves, the system is working well. The paper’s novelty is not just descriptive; it also offers a conceptual response in the form of sociotechnical immunity and a framework that combines early-warning signals, containment actions, and recovery routines. That makes the contribution more than a critique: it is a design-oriented intervention for organizations that want AI benefits without silently hollowing out human capability. The validation, however, is best read as qualitative and framework-oriented rather than as a broad empirical test. The evidence indicates a year-long study in a high-stakes workplace and a cross-domain walkthrough that includes software engineering, which is useful for showing transferability, but it does not establish general prevalence or causal magnitude across sectors. So the paper is strongest as a foundational CHI contribution: it reframes the problem, names a previously under-discussed harm pattern, and proposes a practical response vocabulary. Its main limitation is scope, not ambition. The work is most convincing in expertise-intensive, high-stakes settings where AI use can mask erosion behind apparent success, and it will need further testing in other organizational contexts and power arrangements before the framework can be treated as broadly general.
What Changed
Canon before
Prior CHI work on AI at work typically emphasizes productivity, augmentation, automation, or visible errors and task outcomes; this paper reframes the problem around hidden erosion of expertise, agency, and identity under successful-looking AI use.
Departure from common sense
The paper argues that AI can look beneficial on the surface while quietly degrading expert judgment and agency, so standard productivity gains may conceal harm rather than signal healthy adoption.
Actual novelty
The paper’s main novelty is the AI-as-Amplifier Paradox plus the staged account of hidden harm—"intuition rust" leading to skill atrophy and identity commoditization—paired with a sociotechnical-immunity framework for response and a cross-domain walkthrough that shows how the framework can be used to sense, contain, and recover from erosion in practice.
Evidence
The evidence supports a qualitative, longitudinal account of hidden AI harms in a high-stakes workplace and a framework proposal evaluated through cross-domain walkthroughs. The paper explicitly claims initial gains masked expert erosion, then develops a response framework with early-warning, containment, and recovery mechanisms, and notes transfer across healthcare and software engineering. The claims are coherent and well aligned with the quoted evidence, though the validation remains primarily qualitative and scenario-based rather than experimental.
“findings came from a single domain. Future work can explore these dynamics in multiple domains (finance, cybersecurity, professional training) and populations (e.g., students during AI literacy acquisition) with varied regulatory environments and AI tools to refine immunity mechanisms. Our longitudinal qualitative approach captured the lived experiences of skill erosion. Future mixed-methods studies can leverage our framework to quantify degradation rates, recovery timelines, and immunity thresholds across populations. Additionally, our case involved teams within single organizations; examining ”
actual novelty · Abstract / Section 5.2 · confidence 0.97
“ Initial operational gains hid “intuition rust”: the gradual dulling of expert judgment. These asymptomatic effects evolved into chronic harms, such as skill atrophy and identity commoditization”
departure from common sense · Abstract / Introduction · confidence 0.96
“Interactive Systems Conference (Pittsburgh, PA, USA) ( DIS ’23 ). Association for Computing Machinery, New York, NY, USA, 623–637. doi:10.1145/3563657.3596018 Digital Library Google Scholar [87] Gayatri Chakravorty Spivak. 2023. Can the subaltern speak? In Imperialism . Routledge, 171–219. Google Scholar [88] Bernd Carsten Stahl, Andreas Andreou, Philip Brey, Tally Hatzakis, Alexey Kirichenko, Kevin Macnish, S Laulhé Shaelou, Andrew Patel, Mark Ryan, and David Wright. 2021. Artificial intelligence for human flourishing–Beyond principles for machine learning. Journal of Business Research 124 (2021),”
limitation · Section 7 Limitations and Future Directions · confidence 0.98
“ • Framework for Dignified Human-AI Interaction: A multi-level framework that operationalizes sociotechnical immunity via early-warning signals, containment actions, and recovery routines; evaluated in the primary domain (healthcare) as well as in a second domain (AI-assisted software engineering) to demonstrate cross-domain tran”
validation scope · Section 5.3.1 and 5.3.3 · confidence 0.95
Limits
Method limits
The study is longitudinal and qualitative, so it is strong for surfacing mechanisms and lived experience but not for estimating prevalence, causal effect sizes, or general performance impact across settings.
Deployment limits
The framework is presented as transferable, but deployment will depend on organizational willingness to surface hidden erosion, support worker power, and adapt safeguards to local workflows and institutional constraints.
Boundary conditions
The paper’s own framing suggests the harms emerge in high-stakes, expertise-intensive work where AI use can appear successful while gradually eroding judgment; applicability may be weaker where tasks are low-stakes, highly standardized, or tightly audited.
Position in field
This positions the paper as a reframing move in CHI’s AI-at-work literature: from productivity-centric augmentation narratives toward worker-centered accounts of dignity, expertise preservation, and hidden sociotechnical harm.