Latest / Safe As Podcast / E28: How language shapes blame in investigations

Transcript

E28: How language shapes blame in investigations · Safe As Podcast

• 0:06
  Does the language used in investigations, for instance Bob
• 0:10
  spilled the chemical rather than the chemical was built lead to
• 0:14
  blame in short circuit learning within organisations?
• 0:18
  Let's find out. Good day everyone.
• 0:20
  I'm Ben Hutchinson and This is Safe as a book US dedicated to
• 0:23
  the thrifty analysis of safety, risk and performance research.
• 0:27
  Visit safetyinsights.org for more research.
• 0:31
  This paper from Krista Vessel explored how language in hinder
• 0:35
  or facilitate organisational learning following events.
• 0:39
  It's titled Agentive language in Accent Investigations.
• 0:42
  Why language matters in learning from events, published 2020 in
• 0:47
  ACS Chemical Health and Safety. This is a discussion paper, so
• 0:52
  there's a lot of content, so the presentation today will be a
• 0:55
  little bit different to usual. People naturally want to know
• 0:59
  who or what was responsible for an action, and particularly
• 1:02
  something undesirable. The assignment of action is
• 1:05
  called agency, where an agent is a personal thing that takes an
• 1:09
  active role or produces A specified effect.
• 1:12
  So our example about Bob spilling the chemical, Bob's the
• 1:16
  agent of the action. This simplistic sentence,
• 1:19
  however, doesn't tell us whether Bob spilled the chemical
• 1:22
  intentionally by accident, or simply near the chemical when
• 1:25
  it's spilled. Despite this, we likely assume
• 1:29
  that the agent of the action acted independently and had some
• 1:33
  sort of free will choice to act. These assumptions influence
• 1:37
  learning from events. The assignment of agency is
• 1:40
  subjective, and it's influenced by a host of things like the
• 1:43
  cultures experience and the language of the observer.
• 1:48
  Notably, linguistic framing of events has been shown to
• 1:51
  directly affect the assignment of guilt, blame, and punishments
• 1:55
  of human actors. The language you've written,
• 1:57
  accent reports play a role in identifying these agents and
• 2:01
  linking them to events. Linguistic shortcuts and biases
• 2:04
  are said to reduce the understanding of context around
• 2:07
  an accident that can artificially stop the learning
• 2:10
  process and lead to a false sense of improved safety.
• 2:14
  Language is one of several influences on beliefs of
• 2:16
  causality, but it does have a strong effect.
• 2:20
  As the paper unpacks. Prior workers suggested that
• 2:23
  behaviour can be attributed to a person's internal
• 2:26
  characteristics or their disposition.
• 2:28
  You know, like personality, abilities, attitude or to
• 2:32
  outside influences. You know the the culture,
• 2:35
  society, norms, peer pressure. When observing others, there's a
• 2:39
  tendency to attribute that person's behaviour to internal
• 2:42
  causes, for example, person's disposition or mental state.
• 2:46
  They're a lazy person. This is called the fundamental
• 2:49
  attribution era, where observers underestimate the external or
• 2:53
  situational influences on another person's behaviour and
• 2:57
  overestimate the individual's traits or attitudes.
• 3:00
  They're a lazy person rather than looking at why they might
• 3:04
  be looking for shortcuts. This effect makes it easier to
• 3:07
  judge a person's negative actions as coming from their own
• 3:10
  volition and disposition. Interestingly, different
• 3:13
  cultures may assign different agency.
• 3:17
  Western societies are more likely to identify people other
• 3:20
  than ourselves as the person with the agency, particularly
• 3:23
  for negative outcomes. When evaluating ourselves
• 3:26
  though, we typically focus on success and we distance
• 3:30
  ourselves from failure, which is referred to as self serving
• 3:33
  attribution bias. So for instance, when something
• 3:36
  bad happens in our life, we attribute other factors, it's
• 3:39
  because of someone else or bad luck, we don't attribute it to
• 3:42
  ourselves. English speakers have also been
• 3:44
  shown to have a agentive bias which could affect accent
• 3:47
  analysis. So English speakers are more
• 3:50
  likely to use agentive descriptions for events as
• 3:53
  compared to some other language speakers.
• 3:56
  Being exposed to undesirable events can trigger a defensive
• 3:59
  attribution bias where people defend themselves from the
• 4:02
  concern of being seen as, you know, the cause or victim of a
• 4:06
  mishap. Hence, in the paper, if we can
• 4:08
  categorise a serious action in some way as someone else's
• 4:12
  fault, it's reassuring because we simply need to assure
• 4:15
  ourselves that we are a different kind of person from
• 4:18
  the victim, or we would have behaved differently in that same
• 4:21
  situation. Researchers highlighted that
• 4:23
  simple linguistic changes to an event's description can affect
• 4:27
  the assignment of agency. But quite in the paper, most of
• 4:31
  our own agentive language variations are invisible to us.
• 4:35
  Language priming like agentive languages.
• 4:38
  For instance, he crashed the car or non agentive the car crashed.
• 4:42
  Having studied in different ways, I can't go through all of
• 4:45
  the findings, but one example is how English speakers can better
• 4:49
  remember situations involving agentive language compared to
• 4:53
  non agentive. For in the paper, placing
• 4:56
  attention on individuals involved in accents can actually
• 4:59
  improve memory for people when they recall in the accident.
• 5:02
  So people more likely remember, yeah, Bob crashed the car rather
• 5:05
  than the car crashed. Another really interesting study
• 5:08
  that was covered in here was examination of over 197,000
• 5:13
  criminal trials in London from between 1684 to 1913 found that
• 5:19
  the use of agentive language rhyming resulted in more guilty
• 5:23
  verdicts. So irrespective of the presented
• 5:26
  particulars and the facts of the case, the way that the
• 5:29
  information was framed around agentive language influenced the
• 5:33
  guilty verdicts. So the language of acts,
• 5:35
  investigations, there is a need for linguistic shortcuts.
• 5:40
  You know, we make things easier for ourselves.
• 5:42
  But this can lead investigations towards binary oppositions where
• 5:45
  your pairs of terms are seen as polar opposites like right and
• 5:48
  wrong, success and failure, good and bad.
• 5:51
  We create these either or situations and they can limit
• 5:54
  learning. So there's success, failure.
• 5:57
  Separation is another binary commonality and safety where the
• 6:01
  language is said to be rooted in theories surrounding simple or
• 6:04
  complicated machine systems. But these machinery systems have
• 6:08
  a limited number of parts that can break and function in
• 6:11
  bimodal manners. You know they can be either on
• 6:13
  or off working or not working, but humans can't.
• 6:16
  Humans act in different and inherently complex dynamic ways.
• 6:20
  So we can really only ever say that a human failed or succeeded
• 6:23
  in hindsight once the outcome of the action is known.
• 6:26
  This machine agentive language is often used to describe human
• 6:30
  action when it really isn't suitable for it, But from the
• 6:33
  research, it's unlikely that a single objective story exists
• 6:37
  around an event. Memories, experiences, and
• 6:40
  language will differ when an event is retold by different
• 6:43
  stakeholders, and written descriptions will be subject to
• 6:47
  different sorts of linguistic biases and shortcuts compared to
• 6:50
  people's own recollections. Another element is active versus
• 6:54
  passive verbs, so small differences in language can have
• 6:57
  a large impact on causal attributions like active verb
• 7:01
  voices. We covered some before, but
• 7:03
  Sarah hit the ball that directs agency to Sarah being the one
• 7:06
  that hit the ball, and this hardens attributions of control
• 7:10
  compared to passive voice like ball was hit.
• 7:13
  This attribution effect might still even remain if the agent's
• 7:16
  actions were presented as an unintentional.
• 7:19
  So Sarah didn't mean to hit the ball can still be seen as more
• 7:23
  in control of the action. So discussing the findings
• 7:26
  again, I've skipped a lot. Investigations may search for
• 7:29
  human failure and play significant weight on human
• 7:32
  agency. Blame can be further influenced
• 7:35
  by the fundamental attribution era where dispositional or
• 7:38
  personality based qualities of people are prioritised over
• 7:41
  other factors. You know, the environment, the
• 7:43
  work pressures, the leadership, all that stuff.
• 7:46
  Lots of other studies have looked to blame.
• 7:49
  Here's some other findings that weren't directly covered here,
• 7:51
  but I've actually covered on my side.
• 7:53
  So one study, these are mostly from healthcare as this concept
• 7:56
  of blame conformity. This is where blame for an
• 7:59
  accident can be shifted between individuals as a result of a
• 8:02
  leading eyewitness statement. So they found that over 1/3 of
• 8:06
  participants who read an eyewitness statement that blame
• 8:09
  the individual for the accident went on to then blame the same
• 8:12
  person for its occurrence. So it's been well established
• 8:15
  that when eyewitnesses encounter post event information that
• 8:19
  contradicts or supplements information contained within a
• 8:22
  witnessed event, you often incorporate this new information
• 8:25
  into their own testimony. But let's say somebody observed
• 8:28
  an incident in the street and then if they read someone else's
• 8:31
  testimony, written testimony, they can then tend to
• 8:34
  incorporate that into their own visual eyewitness event.
• 8:38
  So this priming is a relevant issue.
• 8:40
  In fact, some investigation techniques actively control the
• 8:44
  flow of information to the investigators to reduce that
• 8:47
  priming effect. They do their best to ensure
• 8:49
  that the investigators aren't biassed by other information
• 8:52
  until they've had their own chance to sort of reflect on the
• 8:55
  information. Another study from Healthcare
• 8:58
  found that when the investigation reports described
• 9:00
  one or more contributing factors, the investigators were
• 9:03
  more likely to direct blame to the individuals.
• 9:07
  And interestingly, this is something I wouldn't have
• 9:09
  expected that reports listing organisational factors had
• 9:13
  higher odds of apportioning blame towards individuals.
• 9:17
  And the authors really sort of suggest that perhaps those
• 9:20
  reporting incidents may still make fundamental attribution
• 9:23
  errors, you know, blaming individuals, despite recognising
• 9:26
  there were systems factors at play.
• 9:28
  So it is a very human thing to blame.
• 9:31
  And the study also found that, you know, related to the
• 9:34
  fundamental attribution error, it's possible that the voices of
• 9:37
  those most directly connected to the events, you know, the
• 9:39
  frontline operators who were involved in the event aren't
• 9:42
  having an active enough presence in the post event learning
• 9:45
  process. And then another healthcare
• 9:47
  study that I covered on my site also found that investigations
• 9:51
  that had higher word counts, you know, there were larger reports,
• 9:54
  all listed a communication related issue, were more likely
• 9:58
  to attribute blame, which sort of seemed contradictory because
• 10:01
  you would think if they had more space that would have more space
• 10:03
  to talk about the context. But this study actually found
• 10:05
  the opposite. The authors conjecture, perhaps
• 10:08
  having more space, you know, larger word count allows more
• 10:12
  opportunity to incorporate assumptions, hypothesis, and
• 10:15
  value judgments, even if the investigator didn't attend it.
• 10:19
  So what can we make of the findings?
• 10:21
  Blame attributions are a very human thing, so this won't
• 10:24
  easily be changed, at least systematically or permanently.
• 10:28
  Let's first dispel the myth that investigations are an objective
• 10:31
  fact finding exercise. They're not.
• 10:34
  They evolve humans. They can't be.
• 10:36
  This isn't a bug though, it's just a feature.
• 10:39
  So investigations are shaped by social factors as they are
• 10:42
  technical, but not always equally.
• 10:44
  Now first, your author argues that changing the assumptions of
• 10:48
  accident causality and the analysis can have a profound
• 10:52
  impact on practises and logics and belief systems.
• 10:55
  So the author says that you have to change the assumptions that
• 10:58
  drive our belief systems, and language can be a very powerful
• 11:02
  vehicle to do this. My own example is take an
• 11:06
  investigation that says something like the operator
• 11:09
  failed to follow the procedure. It leaves me thinking, and
• 11:13
  that's not a finding, that's the start of the sentence.
• 11:16
  Why didn't they follow the procedure?
• 11:17
  Was it known? Was it tested?
• 11:19
  Was it actually useful to them? How's the task normally done by
• 11:23
  that person or by their peers? Is the procedure normally used?
• 11:27
  Why wasn't it used this time? What was different?
• 11:30
  What makes following the procedure difficult?
• 11:33
  Using the word failed over just reporting that they did not use
• 11:36
  the procedure could be seen as injecting an unnecessary value
• 11:40
  judgement. Indeed, saying the operator
• 11:43
  failed to follow the procedure is itself an example of agentive
• 11:47
  language over something like the procedure was not followed.
• 11:51
  This isn't to justify workarounds in procedural
• 11:54
  departures, it's about using those as critical insights about
• 11:57
  the mismatches between prescribed systems and actual
• 12:01
  practise. Also blame and attributions of
• 12:04
  error masqueraders substantive findings when they rarely are.
• 12:08
  Even things like situational awareness or complacency,
• 12:11
  according to Holnagle, can take on the appearance of folk
• 12:14
  models. These are ideas which purport to
• 12:17
  explain a real world phenomenon, but don't really when you
• 12:20
  scratch below the surface. The author also discusses one
• 12:23
  approach that's been used successfully by them is the
• 12:26
  learning team, and they say that this can help map the
• 12:28
  interrelations between systems factors.
• 12:31
  But of course, learning teams are still susceptible to the
• 12:34
  same very human heuristics and language.
• 12:37
  That's it on SAFE as I'm Ben Hutchinson, please help share,
• 12:41
  rate and review and checkoutsafetyinsights.org for
• 12:45
  more research. Finally, feel free to support
• 12:49
  Safe AS by shouting a coffee link in the show notes.