Latest / Safe As Podcast / E77: Behavior-based safety (BBS) - a balanced review of the science

Transcript

E77: Behavior-based safety (BBS) - a balanced review of the science · Safe As Podcast

• 0:00
  What are your thoughts when I say behaviour based safety
• 0:03
  across the safety literature? Behaviour Based Safety or BBS is
• 0:07
  often presented as a practical way to improve outcomes.
• 0:10
  It's fair to say it's a divisive topic between behavioural
• 0:14
  proponents and those from the so called New View or HOP
• 0:16
  philosophies and more. Instead of trying to resolve
• 0:18
  this debate, I can't, but I want to provide a fairly balanced
• 0:22
  overview of the state of science and some of the key arguments in
• 0:26
  support of and challenging behavioural logics.
• 0:29
  To do this, I've cherry picked multiple studies of the hundreds
• 0:33
  of valuable studies. Anyway, good day everyone,
• 0:36
  welcome to Safe as I'm Doctor Ben Hutchinson.
• 0:38
  Several of the studies we'll cover today highlight both the
• 0:41
  potential benefits and some of the mythological, organisational
• 0:46
  and ethical constraints that shape where the BBS helps or
• 0:50
  simply shifts attention to fixing people, rather than the
• 0:53
  organisational factors that create the risk to begin with.
• 0:56
  First, what is BBS? At its core, BBS is a structured
• 1:01
  way of changing the way work is done by focusing on specific
• 1:04
  behaviours. It involves identifying critical
• 1:07
  observable behaviours linked to injury risk, establishing A
• 1:11
  baseline and then having trained observers.
• 1:13
  Often workers themselves observe work and then give supportive
• 1:17
  non punitive feedback. Those observations are
• 1:19
  aggregated and trended. Not to blame individuals, but to
• 1:22
  learn where conditions, tools or systems are shaping behaviour.
• 1:26
  And as argued by some, when done well, it's somewhat of an
• 1:30
  administrative control designed to complement, not replace,
• 1:33
  engineering and systems level safety controls.
• 1:36
  So study one from Hopkins. What are we to make of safe
• 1:39
  behaviour programmes published in National Research Centre for
• 1:43
  OHS Regulation at the ANU? Hopkins starts with a really
• 1:46
  apartment reflection. Many safe behaviour programmes
• 1:49
  rely on frontline workers observing colleagues and feeding
• 1:53
  back what they see, sometimes one on one and sometimes via
• 1:56
  accounts and trend reports. His core argument is that these
• 2:00
  programmes can fall into what he calls the fallacy of mono
• 2:04
  causality, which is treating unsafe behaviour as the cause
• 2:08
  worth addressing when unsafe behaviour is often the last link
• 2:13
  in a longer causal chain. Instead he unpacks the multi
• 2:16
  causal thinking by walking backwards into the causal chain
• 2:20
  from an unsafe act to the upstream drivers like design
• 2:24
  choices, production pressures and management decisions,
• 2:28
  arguing that once you ask why the unsafe act occurred, causal
• 2:32
  chains often implicate management system conditions.
• 2:35
  And my own interjection. James Reason has this rather
• 2:38
  beautiful reflection on the role of people in large, complex
• 2:42
  failures. That is, rather than being the
• 2:44
  main instigators of an accident, operators tend to be the
• 2:48
  inheritors of system defects created by poor design,
• 2:51
  incorrect installation and bad management decisions.
• 2:54
  Hopkins also emphasises the hierarchy of controls.
• 2:57
  Behavioural programmes often target the lower end, like rules
• 3:01
  or PPE, which are less reliable than eliminating or engineering
• 3:05
  out hazards. In barrier, energy and critical
• 3:07
  control frameworks, there are specific criteria for what
• 3:10
  constitutes human intervention. I don't have time to unpack
• 3:13
  these, but check out my site and the podcasts as I've covered
• 3:16
  these multiple times. Nevertheless, a key insight is
• 3:19
  that these programmes may systematically miss rare but
• 3:23
  critical behaviours like particular workarounds during
• 3:26
  breakdowns or shutdowns, and they might misbehaviors that are
• 3:29
  not observable to the naked eye, such as like permit to work
• 3:32
  failures or isolation error. Permit to work problems in the
• 3:35
  Piper Alpha disaster might be a really good example.
• 3:38
  An absence of bad news about the process was interpreted as the
• 3:42
  system is working. Hopkins argues that frontline
• 3:45
  focused approaches also can't address accidents driven by
• 3:49
  latent conditions where frontline acts are not the
• 3:52
  immediate proximal causes. Finally, Hopkins says
• 3:55
  effectiveness is conditional. These programmes sometimes work
• 3:59
  and sometimes they don't, and a major differentiator is trust
• 4:02
  between the workforce and management, where distrust
• 4:05
  dominates. He argues employers should
• 4:07
  repair underlying issues before expecting behavioural programmes
• 4:11
  to succeed. The next study is from Speaker
• 4:13
  ET al 2022 titled Behaviour based Safety, Today's evidence,
• 4:18
  Journal of Organisational Behaviour Management.
• 4:20
  This paper has two goals, evaluate whether BBS reduces
• 4:24
  injuries and improves safety culture, whatever that is, and
• 4:27
  identify which design components strengthen BBS effectiveness.
• 4:31
  Using data from 88 sites and over 1.3 million observations,
• 4:35
  all implemented via commercial programme IG, the authors report
• 4:39
  modelled injury reductions avering 25% after year one, 34%
• 4:44
  by year 2 and 42% after the third year.
• 4:47
  So, according to their data, BBS programmes are effective for
• 4:51
  these types of measures. Their core findings challenge
• 4:53
  some common assumptions. They report that the greatest
• 4:56
  injury reduction occurred at a moderate observation rate, about
• 4:59
  16 to 65 observations per 100 employees per month, suggesting
• 5:04
  that more observations aren't necessarily better in this data
• 5:07
  set. They report that higher
• 5:09
  percentages of employers acting as observers were associated
• 5:13
  with less improvement in their data, well noting constraints
• 5:16
  about how high participation actually was across their
• 5:20
  sample. Further report that rotating
• 5:23
  observers, rather than keeping the same people in the role
• 5:26
  indefinitely, is more effective, and their guideline table states
• 5:29
  rotation typically every three to 12 months is probably
• 5:32
  appropriate. Finally, they find observations
• 5:35
  were more effective when observers focus on tasks in
• 5:39
  their own area rather than crossing boundaries for a fresh
• 5:42
  pair of eyes, and that using a small number of structured
• 5:45
  comments, sort of flags, like one or two flags, to capture
• 5:49
  context appears more effective than using none or multiple
• 5:52
  flags or prompts. Linking these findings back to
• 5:54
  Hopkins, Hopkins warned that focusing on readily observable
• 5:59
  behaviours can miss deeper hazards speaking ET L.
• 6:03
  This study primarily addresses by emphasising quality in
• 6:06
  context, like task familiarity and contextual flags, rather
• 6:10
  than just the volume of the inspection.
• 6:12
  But BBS isn't a single stable, monolithic concept across many
• 6:16
  decades. To understand why BBS looks
• 6:19
  different across multiple organisations and authors, the
• 6:22
  next paper provides A conceptual history.
• 6:24
  So third is from John Christophe Lacos in 2026 titled The Untold
• 6:30
  Story of Behaviour Based Safety. This is from the extended
• 6:33
  version of safety science. The Coast argues that despite
• 6:36
  BBS's influence and controversy, there hasn't really been any
• 6:40
  published conceptual and critical history of the
• 6:42
  tradition. This paper really aimed to
• 6:44
  remedy that by tracing key texts from the 1970s up until now.
• 6:49
  Methodologically, the paper reports analysing around 20
• 6:52
  books and something like 30 articles, spending both academic
• 6:55
  and consultancy work. And you find but consultancy
• 6:58
  published articles a lot in BBS. The central conclusion is that
• 7:01
  BBS is not monolithic. Well, early BBS drew from
• 7:04
  Skinner influence behaviourism. Skinner been a behaviourist
• 7:07
  psychologist whose operant conditioning treats behaviour is
• 7:11
  shaped by its environment, its consequences which like
• 7:14
  reinforcement, punishment, learning.
• 7:16
  If you want a dog or a person to shake then you keep fitting
• 7:19
  biscuits. But importantly, later
• 7:21
  developments diverged into 3 orientations in the BBS field,
• 7:25
  one the classical traditional which is associated with Salsa,
• 7:28
  2 the theoretical inspirational which is associated with people
• 7:32
  like Geller, and three, organisational critical which is
• 7:36
  associated with Kraus. The Coast also summarises an
• 7:39
  early rationale and principles set from BBS, including a 10
• 7:43
  step style programme model, but emphasises that later authors
• 7:47
  reinterpreted the role of management, organisational
• 7:50
  conditions and the behaviour accent relationship in different
• 7:54
  ways. As highlighted by the likes of
• 7:56
  Scott Galla and Tom Krauss who are non BBS traditionalists,
• 8:01
  behaviour is a result of leadership, culture, system and
• 8:05
  facility design not being the central driver itself.
• 8:07
  That is, behaviour isn't the end into itself.
• 8:10
  So, as mentioned in the crazy article, the starting point for
• 8:13
  improving safety lies not in behaviours but the organisation.
• 8:18
  And importantly, despite critics pointing out some questionable
• 8:22
  and maybe myopic applications of behavioural safety, it's argued
• 8:26
  that the most visible authors producing the main practical and
• 8:29
  academic BBS texts and analysing this study never fully advocate
• 8:34
  such myopic mechanistic applications of behaviour.
• 8:37
  Actually all of them promote systems views in some sense,
• 8:40
  leadership, management, context, the environment.
• 8:44
  So most of these BBS authors support a proactive, positive
• 8:48
  and non punitive view of safe performance.
• 8:50
  And I urge people check out some of Scott Geller work.
• 8:53
  You might forget you're not reading a hop inspired text and
• 8:56
  indeed it's hollowed by Kraus. The popularity of BBS led to
• 9:00
  inconsistent practises on the safety market by a variety of
• 9:04
  poor consulting or half digested rationale and principled by
• 9:08
  different companies, which could happen with any concept really,
• 9:11
  but this led to a lot of cynicism and caution towards
• 9:13
  BBS. 4th is from Cara ET al 2024 titled Implementation of
• 9:19
  Behaviour Based Safety in the Workplace, a review of
• 9:22
  conceptual and empirical literature published in
• 9:24
  Sustainability. Cara and colleagues in their
• 9:26
  literature review present BBS is increasingly implemented and
• 9:31
  argue that the boundaries between standard BBS and related
• 9:35
  interventions can be quite blurry.
• 9:37
  They reviewed 230 articles from the 70s up to about 2023.
• 9:41
  They describe typical BBS implementation steps as a
• 9:44
  baseline analysis intervention, which is like feedback supported
• 9:48
  by observation, post intervention analysis using
• 9:51
  KPIs, and sometimes a withdrawal or reversal phase where you take
• 9:56
  the intervention away and see if behaviour goes back or regresses
• 9:59
  and a few other steps. They also outline foundations in
• 10:02
  behavioural psychology and the ABC model and describe
• 10:05
  historical evolution from supervisory observation
• 10:08
  approaches to more employee LED approaches and later cultural
• 10:12
  partnership approaches. In their synthesis of
• 10:15
  SuccessFactors, they repeatedly highlight the importance of
• 10:18
  trust and communication, worker involvement, active manager or
• 10:23
  supervisor support, and the role of safety.
• 10:25
  Cultural climate, not surprisingly stable findings
• 10:28
  across most domains of safety research, including hot
• 10:31
  philosophies and many others, also find the same things.
• 10:35
  They also describe recurring criticisms in the
• 10:38
  implementation, including privacy concerns, the risk of
• 10:41
  blaming workers or overburdening them, as well as under reporting
• 10:45
  in the challenge of rare high consequence risks.
• 10:47
  So these issues are well known to BBS proponents.
• 10:50
  My interjection, This is a common counter argument from
• 10:53
  behavioural proponents that the core issues of BBS aren't
• 10:57
  structural to behavioural safety, they're more related to
• 11:00
  the design, the implementation and the resourcing of BBS
• 11:04
  approaches or more specifically their misuse.
• 11:07
  Finally, the stated outlook is that future BBS variants will
• 11:10
  integrate traditional observation, positive feedback
• 11:13
  with technology and broader organisational, involved and
• 11:16
  attention to worker and workplace differences.
• 11:19
  So more of that contemporary systems view if this study is
• 11:21
  from Tonsils ET al 2006 titled Effectiveness of Behaviour based
• 11:26
  Safety Interventions to Reduce Accidents and injuries in
• 11:29
  Workplaces published in Theoretical Issues in Economic
• 11:33
  Science and this was a meta analysis.
• 11:34
  Hansel and colleagues said to do 2 things.
• 11:37
  They undertook a meta analysis of the BBS studies as of 2006
• 11:41
  and look at the effects on accident and injury outcomes.
• 11:45
  So since it's an older study, obviously it's missed a lot of
• 11:47
  research until now. For their results, they state 8
• 11:50
  studies achieved statistically significant accident or injury
• 11:53
  reductions, and they report an overall effect size estimate
• 11:57
  indicating statistical significance, while also noting
• 12:00
  there's a lot of differences in the studies and how they
• 12:03
  measured these sorts of improvements and their PBS
• 12:05
  interventions. So they're quite different, and
• 12:07
  these differences make direct comparison a bit more
• 12:10
  challenging. Crucially, they warn that
• 12:12
  statistical significance should be interpreted cautiously
• 12:16
  because most of the studies lacked comparison groups,
• 12:18
  raising the possibility that the observed reductions could be due
• 12:22
  to other factors. Things like regression to the
• 12:23
  main Hawthorne effect, random chance, and more could be one of
• 12:27
  the critical determinants of the findings.
• 12:30
  They also note that there are known reporting distortions, so
• 12:33
  minor injuries may go unreported when monetary rewards depend on
• 12:37
  months without accident. My interjection studies from
• 12:40
  probes suggest that up to 80% of incidents are unreported, so we
• 12:45
  have very little reason to trust reported incident injury data,
• 12:49
  at least for the most minor potential events.
• 12:52
  Nevertheless, their future directions section calls for
• 12:54
  stronger designs. Workplace need better
• 12:57
  assessments, comparable control groups, better reporting of
• 13:00
  sample size and participation rates, accounting for
• 13:03
  confounders, and investigating validity and reliability of
• 13:07
  measurement. So while this paper did find
• 13:09
  overall benefits to BBS, they do caution that because of the weak
• 13:13
  study designs. This really complements Hopkins
• 13:15
  critique that focusing on behaviour can maybe obscure some
• 13:19
  upstream causes because weak designs can't rule out those
• 13:22
  upstream drivers is alternate explanations.
• 13:24
  It also complements Cara's observation that long term
• 13:27
  effectiveness and measurement quality remain recurring issues.
• 13:31
  But I'll come back to this discussion around the state of
• 13:34
  the evidence now. A 6th bonus article that throws
• 13:37
  a cat amongst the pigeons. It's an article from Sydney
• 13:40
  Decker on the safetydifferently.com side
• 13:42
  called the original Hearts and Minds campaign and the
• 13:45
  dereliction of Behaviour based safety.
• 13:47
  It's not peer reviewed of course.
• 13:48
  Decker argues measurement pressure distorts reality.
• 13:52
  He uses the original Hearts and Minds campaign to argue that
• 13:55
  what gets measured can be manipulated, especially when
• 13:58
  leaders demand simple progress graphs and simple measures.
• 14:01
  He argues that the Hearts and Minds label is often used to
• 14:04
  justify behavioural programmes that redirect attention away
• 14:09
  from leadership and towards workers as the problem.
• 14:13
  He describes typical BBS as defining correct behaviours,
• 14:17
  training, measuring compliance and rewarding compliance, which
• 14:20
  might include elements of surveillance and coaching.
• 14:23
  He argues that we shouldn't target behaviour per SE, we
• 14:26
  should wherever possible, target the conditions under which
• 14:29
  behaviour occurs, aim change at whatever is responsible for
• 14:32
  creating those conditions. But importantly, if you watch my
• 14:35
  recent YouTube interview with Decker, you can see broadly
• 14:38
  supports the view from some prominent BBS authors like
• 14:42
  Geller and Tom Krauss that behaviour should be understood
• 14:45
  wherever possible, first from the organisational contextual
• 14:48
  factors, rather than simply as the end to itself.
• 14:51
  And the 7th and final bonus article from Williams and
• 14:55
  Roberts 2018 titled A Holistic approach to improving safety
• 14:59
  performance Publishing professional safety.
• 15:01
  The authors argue that BBS and U View sort of HOP are compatible
• 15:06
  and should be treated as complementary, not competing
• 15:08
  approaches. Both can support safe work
• 15:11
  practises and safety system improvements.
• 15:13
  They emphasise that system factors like faulty equipment
• 15:16
  production pressure over time drive much of the risky
• 15:19
  behaviour or human error and resulting injuries.
• 15:23
  Therefore the first question after an injury should be where
• 15:26
  did the system fail? They outlined hot tools that
• 15:29
  operationalized learning and error prevention before or after
• 15:32
  the work. For instance pre job briefings,
• 15:35
  2 minute analysis, self or peer checking timeouts, post job
• 15:39
  briefings etc. And this can be used to identify
• 15:42
  error traps and error likely situations and strengthen the
• 15:45
  risk control defences. They frame BBS as identifying
• 15:48
  critical behaviours, analysing why risk behaviour occurs, then
• 15:52
  using supportive, corrective and barrier removal system
• 15:55
  modification techniques. They stress limited blame and
• 15:58
  non punitive measures should be used as part of the observation
• 16:02
  checklist and then using trend data to drive system fixes.
• 16:06
  But they also caution that organisations often jump between
• 16:09
  programmes like the flavour of the month, creating siloed
• 16:11
  initiatives and fatigue. Instead we should integrate the
• 16:15
  best elements of all these approaches into one coherent
• 16:18
  system. Some useful takeaways from this
• 16:20
  Williams and Roberts article was we should stop treating BBS and
• 16:23
  HOP as binary choices and mature BBS approaches are broader than
• 16:28
  just checklist and behaviour. It includes the leadership
• 16:30
  behaviours, system improvements, person states, communication,
• 16:34
  not just observing people and I I do wonder, does this involve
• 16:38
  observing the leaders if they're the ones that create most of
• 16:41
  these risks to begin with? Do behavioural observations need
• 16:45
  to check and sit into board meetings?
• 16:47
  If you read some of Scott Geller's work like I mentioned
• 16:49
  before, you might be surprised if you're from this hop element
• 16:51
  just how much alignment there is.
• 16:53
  They also suggest human error is a signal, not a conclusion.
• 16:56
  Hop framing described here treats error as the starting
• 16:59
  point for analysing system weaknesses and strengthen
• 17:01
  defences. Obviously non punitive and
• 17:03
  limited blame and respectful communication are presented as
• 17:06
  essential features for any approach.
• 17:09
  Again, avoid the flavour of the month.
• 17:11
  They warn that cycling new initiatives without integration
• 17:14
  creates scepticism and confusion, which is really hard
• 17:17
  to the uptake of any safety philosophy.
• 17:19
  OK, that's a lot of data. And before I move into some
• 17:22
  takeaways from the articles, I want to reflect on my own
• 17:25
  thoughts. I think there's actually more in
• 17:27
  common between BBS and some of these new viewer hot
• 17:30
  philosophies. And some of us want to admit now
• 17:33
  we might have different units of analysis like BBS is interested
• 17:37
  in behaviour, or they might use some very mechanistic terms like
• 17:39
  unsafe or safe behaviour, which depends on the observer.
• 17:42
  But fundamentally we're all interested in work.
• 17:45
  When you look at modern human factors applications of task
• 17:48
  analysis, critical control verifications under ICMM and the
• 17:52
  like, or construction hazard and control verification approaches
• 17:55
  like the TR method, I think you can begin to see the overlaps
• 17:58
  and complementary natures. We're interested in work, but we
• 18:01
  have different units of analysis to describe it.
• 18:03
  My thoughts? While I've never been a strong
• 18:05
  proponent of BBS itself, I don't see why we'd want to limit our
• 18:08
  options, our adaptive tool boxes, for an approach that,
• 18:12
  regardless of your worldview and political orientations, does
• 18:15
  have a substantial body of evidence behind it, limitations
• 18:19
  included. This doesn't mean every idea and
• 18:21
  concept should be given equal weight, because some concepts
• 18:24
  just have poor evidence behind them.
• 18:25
  But on the other side, we also can't ignore the large empirical
• 18:29
  and industrial tests and uses of BBS.
• 18:32
  Can we trust the injury measures used in these studies and
• 18:35
  articles? Maybe, maybe not.
• 18:38
  That depends entirely on the individual studies and the
• 18:40
  pooled effects, etcetera. And recognising the generally
• 18:42
  consistent findings that the overall body of BBS research is
• 18:46
  relatively weak in methodology limits of the studies, I still
• 18:49
  think it's hard to argue that the burden of evidence hasn't
• 18:51
  shifted from those that promote BBS shifted to those that reject
• 18:55
  it. That is, while critical
• 18:57
  challenges exist and are valid to the validity of BBS research
• 19:02
  with all research and bodies of evidence, I think the onus has
• 19:05
  actually shifted to the critics to demonstrate these approaches
• 19:08
  aren't achieving their core goals or are resulting in
• 19:11
  significant unintended consequences.
• 19:13
  As many arguing, are BBS approaches really shifting the
• 19:16
  focus to minor potential hazards away from process and high
• 19:19
  energy hazards? Are they driving reporting
• 19:21
  underground? Are they impacting trust and
• 19:23
  reporting? I think we're at the point where
• 19:25
  these challenges are interesting, but we need
• 19:27
  empirical evidence to back them up.
• 19:29
  Nevertheless, Roman and authors like Gallow and Krauss have been
• 19:33
  promoting A proactive system focus for decades.
• 19:36
  So we can't group all BBS together, nor should we probably
• 19:40
  group it into the same basket as some of the more mechanistic
• 19:43
  consulting applications, which, mind you, applies for HOP.
• 19:46
  Also, there's still mechanistic applications of HOP.
• 19:48
  So what can we make of the findings?
• 19:50
  Overall, BBS effectiveness is scientifically plausible, but
• 19:54
  the impacts are conditional. Of course, large data sets and
• 19:57
  meta analysis do report injury and incident reductions, but
• 20:01
  multiple sources stress that the results depend on the context
• 20:04
  design and the study quality. Overall, the strength of the
• 20:08
  studies are relatively weak, but.
• 20:10
  We have enough of them to give us a general direction.
• 20:12
  Methodological rigour is a recurring gap across the
• 20:14
  research. So for instance tonsil stress
• 20:17
  poor to marginal study quality and call for stronger controls
• 20:20
  and measurement validity. Noting that was from 2006 and as
• 20:23
  with other research, the long term effects of behavioural
• 20:26
  interventions are relatively limited.
• 20:27
  But it does exist. And importantly, Dominic Cooper
• 20:30
  has cautioned that BBS was never meant as a process safety
• 20:34
  intervention, it was intentionally personal safety.
• 20:37
  So some of the criticism around avoiding major hazards may not
• 20:40
  be fair. The code shows BBS has multiple
• 20:43
  orientations which help explain why BBS can look different
• 20:46
  across different organisations decades.
• 20:48
  We should design BBS so it doesn't substitute for higher
• 20:51
  order controls. We should run alongside hazard
• 20:54
  elimination engineering and high energy applications.
• 20:57
  If you're going to use it, we should critically challenge
• 20:59
  whether you really have widespread behavioural issues or
• 21:03
  whether you have an organisational issue manifesting
• 21:05
  through behaviour. As argued by Decker, human error
• 21:08
  problems might be more organisational issues, and then
• 21:11
  the importance of trust and restore of culture conditions
• 21:13
  matter. Hopkins and Cara and Decker
• 21:15
  explicitly elevate trust and warn about blame dynamics.
• 21:19
  We should combine hop tools and BBS observation and feedback.
• 21:23
  So you capture context, identify error traps, close the loop with
• 21:28
  learning and feedback. I think the critical control
• 21:30
  verification artefacts are probably a sound place to do a
• 21:33
  lot of this. And previously I mentioned the
• 21:35
  TR method which is really useful too if you use BBS and maybe
• 21:39
  reflect on the cards or the observations you use, maybe to
• 21:41
  have more systems and condition factors and prompts about tools,
• 21:45
  equipment, availability, production pressure, leadership
• 21:49
  and other conditions. So observations drive system
• 21:52
  fixes, not just behaviour correction.
• 21:54
  Beware what's missed. Hopkins highlights blind spots
• 21:58
  which are rare critical behaviours and latent conditions
• 22:01
  which may be invisible to behavioural observations.
• 22:04
  We have to make sure to protect psychological safety and
• 22:06
  interpersonal risk taking like observation feedback systems,
• 22:09
  explicitly non punitive and watch for privacy concerns and
• 22:13
  under reporting incentives. We need to prioritise the
• 22:16
  quality of observations and task analysis, task emiliarity,
• 22:19
  meaningful feedback. And this needs to be prioritised
• 22:22
  over quotas and paperwork. Build mechanisms to surface non
• 22:25
  obvious and rare high consequence risk, which could be
• 22:29
  things like permit to work systems, shutdowns and other
• 22:33
  latent factors in the organisation.
• 22:35
  And we have to move away just from the visible PPE and
• 22:37
  housekeeping issues. Thankfully that's it on safe as
• 22:41
  please like, subscribe, comment and share directly on YouTube.
• 22:45
  Leave a comment, it really helps and links in the show notes if
• 22:47
  you'd like to support the channel.