The research design was deliberately structured. Basaran coded the 18 documents with a 28-field codebook covering framing of disruption, responsibility allocation across individual, organizational and policy actors, adaptation resources, psychosocial risks, worker voice and equity dimensions. Every consequential coding decision was linked to a quotation or paraphrase with a page or section reference, creating a page-level audit trail. Two separate coding passes over the entire corpus, both conducted by the author at different points in time, provided an intra-coder consistency check. For the gap variable, the passes agreed on 17 of 18 documents, with a Cohen’s kappa of 0.88. For the responsibility balance, agreement was lower at 10 of 18 documents, though both passes classified all 18 as non-individual-dominant, a finding the author notes runs counter to the initial expectation, informed by critical scholarship, that institutional documents would tend to individualize responsibility.

The documents themselves portray AI-driven disruption as structural rather than episodic. Task-based analyses cited in the corpus, including work by Acemoglu and Restrepo, Felten and colleagues, and Eloundou and colleagues, describe people and AI systems exchanging activities across jobs rather than AI replacing entire jobs wholesale. Evidence from an online labor market indicates that the release of generative AI tools was followed within months by fewer jobs and lower earnings for freelancers in exposed occupations. Because new AI capabilities continue to be introduced, adaptation is required repeatedly rather than once. The corpus also emphasizes that the impact is socio-technical: whether AI supports autonomy or increases workload depends on how it is integrated into work, with electronic performance monitoring producing different effects depending on whether it is designed to support employees or to conduct surveillance.

Across the corpus, 17 of the 18 documents framed the disruption as mixed, acknowledging potential for both improvement and deterioration in job quality, with outcomes presented as contingent on design and governance choices rather than technologically determined. Only one document, an OECD working paper on organizational adoption capacity among small and medium-sized enterprises, framed digitalization primarily as an opportunity. This contingent framing has a direct implication: because the effects of technology are presented as undetermined, the documents identify choices as the factors shaping those effects, and responsibility can be assigned for those choices. The corpus therefore establishes a continuing need for adaptation while also identifying who is expected to undertake it, making it possible to examine responsibility empirically rather than assume it rests with workers.

The allocation of responsibility proved strikingly uneven. Policy and system responsibility was classified as central in 15 of the 18 documents, while organizational responsibility was central in 6 and moderate in 8. Individual responsibility was coded as absent in 2 documents, minor in 14 and moderate in only 2. Governments or policymakers were identified as the actor primarily responsible for change in 10 documents, followed by shared responsibility in 6 and organizations or employers in 2. None of the documents identified workers as the group primarily responsible for change. The overall balance was classified as systemic in 8 documents, shared in 7, organizational in 2 and unclear in 1, with no document classified as individual-dominant in either coding pass.

Yet the study’s central finding is that responsibility has not been transferred to individuals, but neither have the unmet needs been redistributed. A clear Resilience Responsibility Gap emerged in 7 of the 18 documents, a possible or limited gap in 8, and no obvious gap in 3. The gaps followed a nonrandom distribution, appearing most frequently in documents that provided clear empirical analyses of technology use in the workplace rather than in those describing what should be done. Documents reporting establishment surveys, case studies, worker surveys or training-participation data showed where the pairings between adaptation demands and resources broke down in practice. The author notes this concentration partly reflects the coding design, since a gap could be coded only where a document reported evidence on implementation, but the pattern is nonetheless consistent with a failure of institutional reach rather than of institutional intent.

Five inductively identified gap subtypes emerged from the analysis, each supported by at least two documents. The Skills Access Gap concerns high learning expectations paired with unequally accessible provision: one OECD policy brief found that one in three vacancies analyzed had high AI exposure, yet only 0.3 to 5.5 percent of the training courses analyzed delivered AI content, with most existing provision targeting AI professionals with above-average prerequisites. The Participation Gap concerns major technological change with limited worker voice: only 24.1 percent of establishments using data-driven management technologies had discussed their possible impacts with the workforce, and 60 percent of workers considered it unlikely that their employer would take employees’ views into account when introducing AI. The Psychosocial Prevention Gap, the Protection Gap and the Organizational Capacity Gap complete the set, addressing respectively the lag between digital demands and preventive response, the weak social protection available to non-standard workers, and the persistent capacity deficits faced by small firms.

Access to resilience resources proved systematically unequal across all 18 documents. Automation primarily affects low-skilled jobs, yet most low-skilled workers do not participate in training, with an eight-percentage-point gap in the EU27. Economically dependent solo self-employed workers, who report the highest job strain and the least autonomy, have the lowest access to on-the-job training at 24 percent compared with 49 percent of employees. Firm size acts as a systematic sorting mechanism, simultaneously influencing technology adoption, access to training and preventive capacity, with 66 percent of employees in large firms covered by some form of well-being training initiative compared with only 42 percent in micro firms. Employment status can limit access to nearly every resource identified, as workers in atypical, temporary, platform-based or sole-trader employment receive less training, have weaker protection coverage, exercise less agency and receive less preventive support. The author restates the central implication: workers may have identical adaptation needs, yet their access to the resources required to meet those needs can differ entirely.

Psychological Capital, the higher-order construct combining hope, self-efficacy, resilience and optimism, did not appear as an explicit construct in any of the 18 documents, although its components, including trust in employers’ use of AI, autonomy, confidence, adaptability, coping and psychological safety, did appear individually. The author interprets this absence as a disciplinary disconnect rather than a deficiency: institutional documents specify many preconditions of adaptation but lack a framework for the psychological resources through which adaptation operates, while psychological research offers well-developed constructs but rarely extends to the institutional conditions that determine whether resources are available at all. Crucially, the study does not conclude that institutions should teach PsyCap to workers, an inference the author identifies as reproducing precisely the tendency of turning a structural mismatch into an objective for worker development.

Instead, the study advances a testable theoretical proposition: psychological capital is structurally conditioned, and its association with sustainable adaptation is expected to be moderated by the configuration of organizational and institutional resources. Hope requires credible pathways, which active labor-market policies and internal career ladders may either provide or withhold. Self-efficacy develops primarily through mastery experiences, which require training that is both accessible and feasible within workers’ time constraints. Optimism requires a credible basis for expecting positive outcomes, which is difficult to maintain when 60 percent of workers consider it unlikely that their views on AI implementation will be taken into account. Under gap conditions, PsyCap may still predict short-term coping but fail to predict sustainable adaptation because coping is being used as a substitute for missing structural resources. The study accordingly reconceptualizes workforce resilience as distributed adaptive capacity, arising from the alignment between adaptation demands and the individual, organizational and institutional resources available to meet them, with the central organizational question shifting from how to make workers more resilient to how adaptation demands, resources and responsibilities should be distributed so that resilience is sustainable rather than compensatory.

Subject of Research: The Resilience Responsibility Gap: how institutions distribute the burden of adaptation under AI-driven structural disruption

Article Title: The Resilience Responsibility Gap: how institutions distribute the burden of adaptation under AI-driven structural disruption

Article References: The Resilience Responsibility Gap: how institutions distribute the burden of adaptation under AI-driven structural disruption. (n.d.). https://doi.org/10.3389/fpsyg.2026.1974636

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1974636

Keywords: Resilience, Responsibility, institutions, distribute, burden, adaptation, under, AI-driven, structural, disruption, scientific research