Latest / Safe As Podcast / E39: How biased are incident investigators?

Transcript

E39: How biased are incident investigators? · Safe As Podcast

• 0:06
  Our ingrained biases, especially the tendency to blame people,
• 0:10
  are silently undermining safety even when we believe we're being
• 0:14
  objective. This isn't just about mistakes,
• 0:18
  but a systematic effect that can really impact organisational
• 0:23
  learning. Good day everyone.
• 0:25
  I'm Ben Hutchinson and This is Safe as a podcast dedicated to
• 0:29
  the thrifty analysis of safety, risk and performance research.
• 0:34
  Visit safetyinsights.org for more research.
• 0:37
  Today's study is from MacLean and Draw 2023, titled Measuring
• 0:43
  Base Rate Bias Error in Workplace Safety Investigators,
• 0:47
  published in the Journal of Safety Research.
• 0:50
  FYI, I'm covering another paper from this same primary author,
• 0:54
  MacLean in the next pod and that one explores the broader
• 0:58
  research on investigation bias. However, for today, this study
• 1:03
  focused on base rate error and workplace safety investigations
• 1:07
  seeking to understand the magnitude of professional safety
• 1:10
  investigators bias to attribute causes to a person more than
• 1:14
  situational factors. What is base rate error?
• 1:19
  It's a cognitive bias where investigators unconsciously let
• 1:23
  their prior beliefs about how that certain events occur, like
• 1:26
  human error, override the specific evidence of a unique
• 1:30
  incident. This study found it leads them
• 1:32
  to disproportionately blame people using other critical
• 1:36
  factors. But importantly, an unfortunate
• 1:39
  byproduct of cognitive research on bias as practitioners
• 1:43
  interpreting it as something really bad.
• 1:46
  But we need to avoid intuition and heuristics, and we need to
• 1:48
  focus on the facts and analytics.
• 1:51
  But this isn't correct most of the time.
• 1:54
  Our heuristics, our shortcuts and instincts, our rules of
• 1:58
  thumb help us navigate a complex world safely and efficiently.
• 2:03
  Bias isn't a dirty word, but we do need to recognise the
• 2:06
  influence of bias. What were the methods?
• 2:10
  Researchers presented 50 undergraduate students and 12
• 2:13
  professional industrial incident investigators from the forestry
• 2:17
  industry with summaries of a workplace event scenarios.
• 2:22
  Participants informed an initial hypothesis about the most likely
• 2:25
  cause, reported their confidence in their judgement, and assessed
• 2:29
  their perceived objectivity. So, providing a bit of
• 2:32
  background on this area, all people are biassed, including
• 2:36
  all investigators. Prior research has shown people,
• 2:39
  not just investigators, are inclined to attribute more
• 2:42
  'cause to individuals involved in an event than the evidence
• 2:46
  would support. Biases may undermine honest
• 2:49
  investigators, good intentions and their hard work.
• 2:52
  But bias isn't just random error, it's the systematic
• 2:56
  deviation from some sort of standard or evidence based
• 3:00
  objective judgement, if that's even ever possible.
• 3:04
  Some sources of biases range from brain architecture and
• 3:07
  constraints, whereas others originate from contextual
• 3:10
  factors that are related to the environment or specific events.
• 3:15
  When assessing events, people naturally compare that event to
• 3:18
  some sort of base rate frequency.
• 3:21
  In other words, it's probability.
• 3:23
  Judgments can be biassed when population base rates, said to
• 3:26
  be the actual frequency of an occurrence, are inconsistent
• 3:29
  with the investigators beliefs, leading them to believe that the
• 3:33
  event was more or less likely to occur compared to the population
• 3:37
  frequency. They call human error bias as
• 3:40
  the cognitive tendency in workplace situations where
• 3:43
  worker action is overemphasised as a causal factor.
• 3:47
  In event, further explicit or implied organisational pressure
• 3:51
  to find human error may guide, even perhaps uncautiously,
• 3:56
  investigators conclusions about the responsibility of the
• 3:59
  worker. And they note that once an
• 4:01
  investigator has a working theory of worker fault,
• 4:05
  confirmation bias demonstrates how his hypothesis can shape
• 4:09
  information collection and interpretation to endorse that
• 4:13
  investigator's predominant view, and more thorough investigation
• 4:17
  may simply entrench that belief with additional confirming
• 4:20
  evidence. So what did they find?
• 4:23
  They found that professional investigators, the very people
• 4:27
  trained to be apparently objective, exhibited a real,
• 4:31
  clear, and consistent human error bias.
• 4:34
  Therefore, there was an illusion of objectivity.
• 4:37
  These same professional investigators simultaneously
• 4:39
  believed they were objective and highly confident in their
• 4:43
  assessments. This highlights a fundamental
• 4:46
  disconnect between their perceived performance and actual
• 4:49
  cognitive processing. While the lay student group also
• 4:53
  showed this bias, the study revealed that this bias was
• 4:56
  significantly larger and more pronounced in the professional
• 5:00
  investigator groups. This suggests that a robust
• 5:04
  human error bias was present across the board, but
• 5:07
  experience, as in greater levels of experience in the
• 5:10
  investigators, rather than mitigating the effect, appeared
• 5:14
  to exacerbate the bias, leading to worker action being
• 5:18
  implicated as the primary causal factor more often than other
• 5:22
  systemic or situational elements.
• 5:25
  So they found some sort of dangerous disconnect, a
• 5:28
  miscalibration where the participants self rated
• 5:31
  confidence and perceived objectivity were entirely
• 5:35
  unrelated to the actual quality or accuracy of their causal
• 5:39
  allocations. In simpler terms, how confident
• 5:42
  investigators felt had no bearing on whether their
• 5:45
  conclusion was actually accurate or not.
• 5:48
  This miscalibration's a key take away.
• 5:51
  It suggests that safety initiatives and corrective
• 5:54
  actions could be, and likely are proceeding more based on the
• 5:58
  persuasiveness in confidence of the investigators delivery
• 6:02
  rather than on the true quality, accuracy or evidence base of
• 6:05
  their findings. This can lead to resources being
• 6:08
  misdirected or underlying factors remaining unaddressed
• 6:13
  and a false sense of security in the workplace.
• 6:16
  In essence, a study demonstrates that the human cognitive biases,
• 6:20
  particularly the tendency to overemphasise individual blame
• 6:24
  and not only pervasive in safety investigations, but can also be
• 6:29
  magnified by experience and dangerously masked by
• 6:33
  investigators misplaced confidence.
• 6:36
  This creates A systemic vulnerability where the very
• 6:38
  process is designed to identify and mitigate hazards, may
• 6:42
  instead entrench misleading conclusions, leaving operational
• 6:45
  risks unaddressed. What can we make of the findings
• 6:49
  for one not considering investigations to be absolute
• 6:53
  and black and white searches for truth then hot and likely never
• 6:58
  will be in almost any practical application.
• 7:01
  The author discusses the role of debiasing techniques.
• 7:04
  To date, the evidence is pretty patchy on debiasing across
• 7:08
  different applications. The author suggests that one
• 7:11
  debiasing technique called considering alternatives may
• 7:14
  help in investigations. This is where alternative and
• 7:18
  plausible hypothesis are generated to contradict the
• 7:21
  initial explanations. Other suggestions include
• 7:24
  reducing the time pressures faced by investigators or
• 7:27
  fatigue or and other constraints.
• 7:30
  It's argued that even the idea of an accurate investigation may
• 7:34
  push investigators more to overanalyze things and
• 7:37
  speculate. Indeed, quote in the paper,
• 7:40
  investigators with a compulsion to be flawless would experience
• 7:43
  increased stress and pressure. So maybe close enough is good
• 7:48
  enough, if we still learn and improve what really matters, and
• 7:53
  maybe working on the underlying blame logics present within
• 7:56
  organisations. That is, how does management and
• 8:00
  others in the organisation perceive failure?
• 8:02
  How do they respond to failure? Are people bad apples that
• 8:05
  introduce risk to otherwise safe systems?
• 8:08
  Or do we see people situated within the overall environment?
• 8:13
  In any case, I'll cover another more systematic review of the
• 8:16
  research and the biassing techniques from that same
• 8:19
  primary author in the next episode.
• 8:22
  What were some limitations? First, it was based on a
• 8:25
  constructed scenario and not a real world investigation per SE.
• 8:29
  Also, the total sample of investigators, just 12, wasn't
• 8:33
  that large, but still really useful data came out of this
• 8:37
  study. That's it.
• 8:39
  On Safe as I'm Ben Hutchinson, please help share, rate and
• 8:43
  review and checkout Safety insights.org for more research.
• 8:48
  Finally, feel free to support Safe As by shouting a coffee
• 8:52
  link in the show notes.