Nine angles of analysis.
Six Gollac families, supplemented by three exploratory axes specific to AI use.
Psychosocial risks · Organisation · Health
AI becomes a risk when work organisation creates exposure.
A model alone does not explain a health effect. Risk develops when the use of AI changes workload, autonomy, relationships, work quality or professional security without adequate resources for regulation and control.
Collective analysisReal workNo individual diagnosis
When the balance breaks
Destruction · 1836
The city’s apparent order collapses when its supports can no longer hold. The work reminds us that, at work, an equilibrium can quickly become fragile when workload, accountability, routes for challenge or skills are not followed with care.
Examine the organisation.
A score, error or recommendation becomes hazardous only within a particular work situation.
Think in terms of exposure.
Duration, frequency, combined constraints, available resources and room for action shape the level of risk.
01 / What creates exposure
Four organisational choices to examine first.
Start not with “is AI dangerous?”, but with who does what, under which targets, with what authority given to the system and what real means to regain control.
Targets raised too quickly.
Theoretical time savings become higher volumes or shorter deadlines before the actual time required for checking is known.
May drive: intensity, interruptions, work spilling beyond hours.Decisions removed from real work.
The tool is selected without involving users, worker representatives or those who will handle errors and difficult cases.
May drive: loss of autonomy, conflict, workarounds.Responsibility without resources.
Workers must approve outputs without enough time, sources or authority to challenge them.
May drive: cognitive overload, anxiety, impossible accountability.Unclear professional change.
Tasks are removed or recombined without training pathways, skill maintenance or credible information about jobs.
May drive: insecurity, disengagement, skill erosion.02 / Gollac framework
The six families of psychosocial risk factors.
These factors describe employment and organisational conditions that may affect health. They are neither symptoms nor diagnoses.
Work intensity and time.
Volume, deadlines, interruptions, hours and complexity.
With AI: raised targets, invisible correction work, permanent alerts.Emotional demands.
Managing one’s emotions, distress and conflict.
With AI: announcing, explaining or repairing a difficult automated decision.Autonomy.
Freedom to choose methods and pace and to use one’s skills.
With AI: imposed workflow, recommendations that are hard to override, increased monitoring.Social relations.
Support, cooperation, recognition, conflict and management quality.
With AI: dispersed accountability, lost exchanges, no effective person to challenge.Conflicts of values.
Being unable to do work consistent with professional standards and meaning.
With AI: speed prioritised over quality, outputs judged unfair or unacceptable.Job insecurity.
Uncertainty about jobs, roles, income and future change.
With AI: unclear task futures, shifting evaluation criteria, continuous reorganisation.03 / Exploratory supplement
Three mechanisms especially relevant to AI.
These axes supplement the Gollac framework in this site’s questionnaire. They structure discussion, but the adaptation is not a validated psychometric instrument.
Opacity and contestability.
Understanding an output well enough to spot errors, request an explanation and obtain effective human review.
Critical signal: an output affects someone without a route able to change the decision.Supervision burden.
Checking, correcting and justifying plausible outputs requires time, sustained attention and available expertise.
Critical signal: checking is mandatory but incompatible with workload and deadlines.Skill erosion.
Automation may reduce practice, junior learning, transmission and the ability to work without the tool.
Critical signal: dependence on the system without organised skill maintenance.04 / Possible consequences
From exposure to health effects.
Effects depend on duration, intensity, combined constraints and available resources. A single symptom cannot establish that AI caused it.
What may appear in work.
- fatigue and concentration difficulties
- irritability, nervousness, helplessness
- sleep disturbance and inadequate recovery
- errors, workarounds and team tension
Effects associated with psychosocial exposure.
- chronic stress, anxiety or depressive disorders
- occupational burnout
- musculoskeletal disorders
- cardiovascular risk and adverse health behaviours
Possible collective consequences.
- absence and staff turnover
- accidents, errors and lower quality
- deteriorating social climate
- skill loss and recruitment difficulties
Methodological point: prevention addresses organisational factors and work situations. It does not measure presumed individual vulnerability or produce personal risk scores.
05 / Act
Four decisions, without adding procedural burden.
These actions link to tools already explained elsewhere on the site.
Describe.
Task, affected groups, authority of the output, constraints and baseline situation.
Limit.
Reversible pilot, excluded uses, right to override, fallback and stopping criteria.
Provide resources.
Checking time, training, sources, support and explicit allocation of responsibility.
Follow up.
Compare workload, autonomy, errors, skills, teams and health before and after.
06 / Sources
Framework and evidence.
The three AI axes are an exploratory editorial supplement; the six factors and health effects are grounded in the institutional sources below.