Human Error
Lecture 1: Human Error Classification, 15-11-2024
Error: The failure of planned actions to achieve their desired goal without some unforeseen
or chance intervention. Error types (Taxonomy):
- Errors of Automation: slips (attention failure), lapses (memory failure)
- Errors of Conscious Control: mistakes (intention failure), violations (people are aware
of violating the law)
Rasmussen - Human Performance
- Knowledge level → bounded rationality & thinking
- Rule level → pattern matching & recognition (if...then...)
- Skill level (highest level) → highly automated, very
little/no effort
Skill level is most energy efficient, so the preferred way of
performance. More skill means more automated behavior and
less effort.
Generic Error-Modelling system (GEMS)
- Skill level → Slips and Lapses
- Rule level → Rule Based mistakes
- Knowledge level → Knowledge Based Mistakes
Errors that typically proceed detection of a problem: Skill based errors (Slips & Lapses)
Errors that typically follow detection of a problem: Rule and Knowledge based Mistakes
Skill Level Errors: Inattention/Omission
- Double-capture slips: distracter captures executive control; stronger automation
takes over (forget something special)
- Omissions following interruptions: attentional check fails after an external event
(interrupted normal behavior)
- Delay reduces control → detached intentions, environmental capture, multiple
sidesteps (what am I doing here?)
- Perceptual confusions: a pattern matching problem
- Interference errors: a control issue, “blends”, “spoonerisms”
o Freudian slip: an error in speech, memory, or physical action that occurs due
to the interference of an unconscious subdued wish or internal train of
thought (calling your teacher mom)
Skill Level Errors: Overattention → omissions, repetitions, reversals
- Sometimes it is just best not to think (too much)
Rule Level Errors: Misapplication of good rules
- First exceptions (over-generalization): the first time an exception is met, leads to the
creation of sub-rules
- Countersigns and nosigns
o Signs: all or some of the conditional aspects of the rule met
, o Countersigns: indicate another rule is appropriate
o Nosigns: do not relate to any existing rule and are simply noise in the system
- Informational overload: complex real-world situations, limited resources to: notice
signs and countersigns/ignore nosigns
- Rule strength: the stronger (more frequently used) a rule the easier it will “fire”,
including partial matches
- General Rules: stronger rules, exceptions are exceptional
- Redundancy: Repeated encounters result in ignoring redundant information,
increasing the strength of the remaining cues
- Rigidity: If it ain't broke don’t fix it
Rule Level Errors: Application of bad rules
- Encoding deficiencies: lack of encoding, incorrect encoding (Naïve Physics)
- Action deficiencies: wrong rules, inelegant rules, inadvisable rules
Knowledge level errors
- Selectivity: attention directed towards psychological salient rather than logically
important
- Workspace limitations: limited resources, recent info bias
- Out of sight out of mind: availability heuristic, ‘A man with a hammer…’
- Confirmation bias: Effort after meaning & expectations
- Overconfidence: Disregard contradictory signs
- Biased reviewing: The “check-off” illusion
- Illusory correlation: Correlation is not causation
- Halo effects: if someone is physically attractive, they must also be a good person
- Problems with causality
o Underestimating future irregularities
o Representativeness heuristic: causality biased on similarity between cause
and effect
o Availability heuristic: anecdotal arguments
o Hindsight bias
o Illusion of control
- Problems with complexity
o Delayed feedback
o Insufficient considerations of processes in time
o Thinking in causal series
o Thematic vagabonding (flit, superficial)
o Encysting (“myopic” detailed attention, ignore larger issues)
o Difficulties with exponential developments
Violations: done with intent to violate, but NOT to have negative consequences. No intention
to do harm. Some violations are routine violations → cutting corners & saving effort
Risk: The probability of an outcome x the consequence of that outcome
Risk Monitoring: Target level of risk? Target feeling of risk/task difficulty/effort? Or zero risk?
Lecture 1: Human Error Classification, 15-11-2024
Error: The failure of planned actions to achieve their desired goal without some unforeseen
or chance intervention. Error types (Taxonomy):
- Errors of Automation: slips (attention failure), lapses (memory failure)
- Errors of Conscious Control: mistakes (intention failure), violations (people are aware
of violating the law)
Rasmussen - Human Performance
- Knowledge level → bounded rationality & thinking
- Rule level → pattern matching & recognition (if...then...)
- Skill level (highest level) → highly automated, very
little/no effort
Skill level is most energy efficient, so the preferred way of
performance. More skill means more automated behavior and
less effort.
Generic Error-Modelling system (GEMS)
- Skill level → Slips and Lapses
- Rule level → Rule Based mistakes
- Knowledge level → Knowledge Based Mistakes
Errors that typically proceed detection of a problem: Skill based errors (Slips & Lapses)
Errors that typically follow detection of a problem: Rule and Knowledge based Mistakes
Skill Level Errors: Inattention/Omission
- Double-capture slips: distracter captures executive control; stronger automation
takes over (forget something special)
- Omissions following interruptions: attentional check fails after an external event
(interrupted normal behavior)
- Delay reduces control → detached intentions, environmental capture, multiple
sidesteps (what am I doing here?)
- Perceptual confusions: a pattern matching problem
- Interference errors: a control issue, “blends”, “spoonerisms”
o Freudian slip: an error in speech, memory, or physical action that occurs due
to the interference of an unconscious subdued wish or internal train of
thought (calling your teacher mom)
Skill Level Errors: Overattention → omissions, repetitions, reversals
- Sometimes it is just best not to think (too much)
Rule Level Errors: Misapplication of good rules
- First exceptions (over-generalization): the first time an exception is met, leads to the
creation of sub-rules
- Countersigns and nosigns
o Signs: all or some of the conditional aspects of the rule met
, o Countersigns: indicate another rule is appropriate
o Nosigns: do not relate to any existing rule and are simply noise in the system
- Informational overload: complex real-world situations, limited resources to: notice
signs and countersigns/ignore nosigns
- Rule strength: the stronger (more frequently used) a rule the easier it will “fire”,
including partial matches
- General Rules: stronger rules, exceptions are exceptional
- Redundancy: Repeated encounters result in ignoring redundant information,
increasing the strength of the remaining cues
- Rigidity: If it ain't broke don’t fix it
Rule Level Errors: Application of bad rules
- Encoding deficiencies: lack of encoding, incorrect encoding (Naïve Physics)
- Action deficiencies: wrong rules, inelegant rules, inadvisable rules
Knowledge level errors
- Selectivity: attention directed towards psychological salient rather than logically
important
- Workspace limitations: limited resources, recent info bias
- Out of sight out of mind: availability heuristic, ‘A man with a hammer…’
- Confirmation bias: Effort after meaning & expectations
- Overconfidence: Disregard contradictory signs
- Biased reviewing: The “check-off” illusion
- Illusory correlation: Correlation is not causation
- Halo effects: if someone is physically attractive, they must also be a good person
- Problems with causality
o Underestimating future irregularities
o Representativeness heuristic: causality biased on similarity between cause
and effect
o Availability heuristic: anecdotal arguments
o Hindsight bias
o Illusion of control
- Problems with complexity
o Delayed feedback
o Insufficient considerations of processes in time
o Thinking in causal series
o Thematic vagabonding (flit, superficial)
o Encysting (“myopic” detailed attention, ignore larger issues)
o Difficulties with exponential developments
Violations: done with intent to violate, but NOT to have negative consequences. No intention
to do harm. Some violations are routine violations → cutting corners & saving effort
Risk: The probability of an outcome x the consequence of that outcome
Risk Monitoring: Target level of risk? Target feeling of risk/task difficulty/effort? Or zero risk?
