Use, Misuse, Disuse, Abuse of Automation

emerging Updated: 2026-06-07
automation supervisory-control levels-of-automation trust complacency overreliance automation-bias false-alarms monitoring human-centred-automation remote-operations

Use, Misuse, Disuse, Abuse of Automation

Status: emerging
Last updated: 2026-06-07
Sources: 001872097778543886.Pdf
Tags: [automation, supervisory-control, levels-of-automation, trust, complacency, overreliance, automation-bias, false-alarms, monitoring, human-centred-automation, remote-operations]

Summary

Parasuraman and Riley (1997) organise the human factors of automation around four patterns of human interaction: use, misuse, disuse, and abuse. Use is the operator's voluntary engagement or disengagement of automation, shaped by trust, workload, perceived risk, self-confidence, and large individual differences. Misuse is overreliance — automation bias and failures to monitor a "strong but silent" system. Disuse is neglect, typically of alerting systems made untrustworthy by false alarms and base-rate neglect. Abuse is the technology-centred application of automation by designers and managers without regard for the consequences for human performance, which itself breeds misuse and disuse. The paper's prescription is human-centred automation: keep the operator informed, involved, and in authority.

Body

Context

Parasuraman and Riley (1997) review theoretical, empirical, and analytical work on human use of automation across aviation, manufacturing, ground transport, and medicine, deliberately not focusing on any single system. They define automation as the execution by a machine of a function previously done by a human, and treat it as a continuum of levels rather than an all-or-none property (PDF pp. 3–4, orig. pp. 231–232). The article is a conceptual anchor for this knowledge base: remote operation centres are a case of supervisory control in which operators decide when to rely on autonomy, monitor it, and intervene — exactly the behaviours the four-part taxonomy describes. It underpins the maritime trust and transparency work in Trust In Human Autonomy Teaming and Human In The Loop Automation Transparency, and it complements Bainbridge's ironies of automation held in the AI knowledge base (see ironies-of-automation).

Key Points

Use is hard to predict. Whether an operator engages automation depends on attitudes, mental workload, cognitive overhead, trust, self-confidence, and risk, and these interact. Attitudes toward automation do not reliably predict behaviour — two studies found no relationship between attitudes and actual reliance (PDF pp. 6–7, orig. pp. 234–235). Workload's influence is equivocal: operators cite high workload as a reason to automate, yet laboratory manipulations of task difficulty often fail to change usage (PDF pp. 7–8, orig. pp. 235–236). Cognitive overhead — the effort of deciding whether and how to use an aid — can deter use even when the aid is beneficial (PDF p. 8, orig. p. 236). Trust is central: operators tend to rely on automation when their trust in it exceeds confidence in their own manual skill, and they may keep relying on it even through failures (PDF pp. 8–9, orig. pp. 236–237). Riley's interactive model gathers these factors, but pervasive individual differences make prediction of any one operator's behaviour difficult (PDF p. 10, orig. p. 238).

Misuse is overreliance. Excessive trust leads operators to rely uncritically on automation and to monitor it poorly. Two mechanisms recur. Automation bias makes the automated cue a heuristic replacement for vigilance, producing omission errors (missing a problem the automation did not flag) and commission errors (following an inappropriate automated directive); expertise does not confer immunity (PDF pp. 11–12, orig. pp. 239–240). Monitoring failures grow with manual task load: in the authors' flight-simulation paradigm, operators detected over 70% of engine malfunctions when monitoring manually but far fewer when the task was automated and they were loaded with other tasks, while detection was near-perfect when monitoring was the only task (PDF pp. 12–13, orig. pp. 240–241). Constant, unchanging automation reliability breeds the most complacency; variable reliability is monitored more closely (PDF p. 13, orig. p. 241). Countermeasures include making automation state indicators salient through display integration and adaptive task allocation — briefly returning a task to manual control restores subsequent monitoring (PDF pp. 14–15, orig. pp. 242–243). High-authority, high-autonomy automation worsens the problem by creating "strong but silent" agents and can reduce situational awareness (PDF p. 15, orig. p. 243).

Disuse is the neglect or disabling of automation, driven largely by false alarms. Setting an alerting system's decision criterion to minimise misses raises the false-alarm rate, and a low base rate of the hazardous condition makes the posterior probability of a true alarm low even for a sensitive detector — so operators who have been "cried wolf" at once too often ignore, slow their response to, or physically disable alarms (PDF pp. 16–18, orig. pp. 244–246). The design lesson is that alerting-system thresholds must be set against the base rate of the condition, not the criterion alone, and that "likelihood alarms" conveying graded danger may earn more trust than binary ones (PDF pp. 17–18, orig. pp. 245–246).

Abuse is automating functions for technical or economic reasons without regard for the operator's resulting role. Automation does not remove human error; it relocates it from the operator to the designer (the weight-on-wheels sensor that blocks the pilot exactly when it fails) and can make automation a surrogate for the manager (a policy refusing manual control in bad weather contributing to a transit-train collision) (PDF pp. 18–19, orig. pp. 246–247). Technology-centred deployment reduces operators to monitors, which promotes overreliance, and demands a proportionately higher level of feedback the more authority the automation holds (PDF pp. 19–20, orig. pp. 247–248).

Conclusion

Parasuraman and Riley (1997) argue that misuse, disuse, and abuse share a root cause: mismatched expectations among designers, managers, and operators about how automation will actually be used. Their conclusions converge on human-centred automation — operators should be taught to make rational use decisions; overreliance and its antecedents should be anticipated and countered with salient feedback; alerting thresholds must account for base rates; and the operator's role should be defined from human responsibilities and capabilities rather than left as a by-product of what was technically convenient to automate. Keeping the operator actively involved and in authority is presented as a net safety benefit even when full automation could perform a function better, because only the human can hold fiduciary responsibility for system safety.

References

Bainbridge, L. (1983) 'Ironies of automation', Automatica, 19(6), pp. 775–779. To be validated.

Lee, J. and Moray, N. (1992) 'Trust, control strategies and allocation of function in human-machine systems', Ergonomics, 35(10), pp. 1243–1270. To be validated.

Parasuraman, R. and Riley, V. (1997) 'Humans and Automation: Use, Misuse, Disuse, Abuse', Human Factors, 39(2), pp. 230–253. doi: 10.1518/001872097778543886. parasuraman1997humans

Riley, V. (1989) 'A general model of mixed-initiative human-machine systems', Proceedings of the Human Factors Society Annual Meeting, 33(2), pp. 124–128. To be validated.

Open Questions

  • The paper predates modern machine learning, yet automation bias, monitoring failure, and base-rate-driven distrust map directly onto today's AI decision aids. How far the four patterns transfer to opaque, adaptive autonomy is worth tracking against newer sources in the corpus.
  • Riley's model captures the factors influencing use but cannot predict an individual operator's choice because of large individual differences. What design or training levers most reduce that unpredictability is left open.
  • The prescription to keep the operator "in authority" can conflict with cases (the Conrail and Gaithersburg accidents cited) where authority was deliberately moved to automation because the operator was judged untrustworthy. The boundary between appropriate operator authority and justified override is unresolved.