Do you see procedures as concrete actions that specify the one correct way of working, or more as resources to shape work and sensitise people to risk?
Let’s unpack the model 1 / model 2 perspective of rules and see which resonates best with workers.
Today’s paper is Hendricks, J. W., & Peres, S. C. (2021). Beyond human error: An empirical study of the safety Model 1 and Model 2 approaches for predicting workers’ behaviors and outcomes with procedures. Safety science, 134, 105016.
(Spoiler: We need both)
Spotify: https://open.spotify.com/episode/4bkXtRa1k4ARsF07L8YyGE?si=8gUX6b2pSBSO9XfvIWlgDg
Make sure to subscribe to Safe As on Spotify/Apple, and if you find it useful then please help share the news, and leave a rating and review on your podcast app.
I also have a Safe As LinkedIn group if you want to stay up to date on releases: https://www.linkedin.com/groups/14717868/?lipi=urn%3Ali%3Apage%3Ad_flagship3_detail_base%3Bhdg8uJYYT%2BmsMqZvpHBmdQ%3D%3D
Shout me a coffee (one-off or monthly recurring)
Transcript:
6.921375 11.641375 Beyond the checklists and compliance lies the human reality of work.
11.641375 16.721375 A more progressive approach argues for a shift in thinking, from controlling behaviour to
16.721375 21.961375 supporting it, reminding us that effective procedures are a tool to empower people, not
21.961375 23.161375 chains that bind them.
23.161375 27.481375 But what do workers and management think about procedures?
27.481375 28.481375 Let’s find out.
28.481375 33.681375 G’day everyone, I’m Ben Hutchinson and this is Safe As, a podcast dedicated to the thrifty
33.681375 37.761375 analysis of safety, risk and performance research.
37.761375 41.041375 Visit safetyinsides.org for more research.
41.041375 48.081375 Today’s study is from Hendricks and Perez, 2021, titled “Beyond Human Error, An Empirical
48.081375 55.281375 Study of the Safety Model 1 and Model 2 Approaches for Predicting Workers’ Poveries and Outcomes
55.281375 56.281375 with Procedures”.
56.281375 61.721375 I covered another paper on procedures from these same authors in an earlier episode.
61.721375 66.681375 This study explored the relationship between individual and system-level variables on procedure
66.681375 74.281375 use and procedural departures by interviewing 174 industry personnel and they also measured
74.281375 77.921375 some other stuff that I won’t talk about right now.
77.921375 83.481375 Individual factors included things like people’s perceptions of the utility of procedures, their
83.481375 86.801375 thoughts around rule compliance and a few other things.
86.801375 91.321375 System variables included the procedural quality, perception of the procedure change
91.321375 94.281375 process and safety climate.
94.281375 99.601375 What was really interesting is they framed the analysis via the Model 1 and Model 2 concept.
99.601375 102.681375 Now just for clarity, this isn’t Safety 1 and Safety 2.
102.681375 107.801375 The concept originally came from Sydney Decker, but it was really fleshed out and discussed
107.801375 113.601375 by Hale and Boris in two 2013 papers, really quickly.
113.601375 118.441375 So based on Decker’s conceptualization and then Hale and Boris’s expansion of it, it
118.441375 122.881375 said that an organization’s approach to procedural systems can be kind of thought about as a
122.881375 125.761375 Model 1 or Model 2 approach.
125.761375 131.521375 Model 1 sees procedures as rules to be followed closely and strictly, therefore deviations
131.521375 135.641375 are always a mistake and reflect a problem with the worker.
135.641375 139.961375 Much of the previous research has really looked for methods to decrease human error,
139.961375 143.681375 trying to ways to fix the worker via rules.
143.681375 148.961375 This Model 1 approach tends to see procedures as things that need to be followed rigidly
148.961375 152.281375 and workers shouldn’t be deviating from the procedures.
152.281375 157.921375 So from the Model 1 perspective, rules are very rationalistic and prescriptive and the
157.921375 162.961375 rules seem to be the one correct way to do the work, there to protect workers from their
162.961375 164.601375 own fallibility.
164.601375 169.681375 These rules are seen as essentially static, to be worked out as a one-off exercise and
169.681375 173.801375 only to be modified when the activity changes substantially.
173.801375 178.641375 The Model 1 is essentially top-down, it gets applied to the workforce and then management
178.641375 181.121375 act as the enforcers.
181.121375 186.521375 In contrast, the Model 2 perspective sees procedures more as tools to support work and
186.521375 191.801375 resources for action, so some prior research found that procedures alone weren’t sufficient
191.801375 196.521375 to support work, but they do often best serve as resources for action.
196.521375 201.161375 The Model 2 has more of a dynamic nature to it, allowing some level of flexibility and
201.161375 206.761375 interpretation of procedural systems, which is then theorised to help maximise the utility
206.761375 207.761375 of a procedure.
207.761375 213.521375 So the Model 2 view of rules sees rules as patterns of behaviour, socially constructed,
213.521375 219.241375 emerging from experience, that characterises local and situated in the activity.
219.241375 224.321375 So Model 2 sees rules essentially bottom up and dynamic, it recognises that rules can
224.321375 229.561375 never be complete, and the written ones are seen as essentially underspecified, requiring
229.561375 235.481375 a process of translation and adaptation by workers before application to any given specific
235.481375 236.481375 situation.
236.481375 241.481375 So, rules are seen in Model 2 as a support and guidance for the expert, as templates
241.481375 243.561375 and resources for adaptation.
243.561375 250.561375 Now, importantly, none of this suggests that Model 1 is bad and Model 2 is always good,
250.561375 252.321375 or vice versa.
252.321375 255.881375 We need both Model 1 and Model 2 thinking with their procedures.
255.881375 261.841375 What’s important is really understanding how they interact and how rules are used and
261.841375 263.641375 perceptualised in the field.
263.641375 269.201375 But one of the key prompts of this study is that improving procedural systems might be
269.201375 274.841375 limited if the focus is placed mostly on trying to tightly control worker behaviour via the
274.841375 280.521375 Model 1 approach, rather than accounting for broader organisational factors.
280.521375 282.401375 So what did they find?
282.401375 286.601375 Evidence was found for both individual and system variables.
286.601375 291.521375 A range of things like perceived procedure quality, role experience, attitudes towards
291.521375 296.081375 the procedure and procedural utility, and people’s own compliance attitudes were all
296.081375 299.601375 related to some types of outcome measures.
299.601375 303.841375 Attitudes towards procedure utility was found to be the only predictor of incidents and
303.841375 309.961375 near misses, where better attitudes regarding the utility of procedures was related to fewer
309.961375 314.481375 self-reported yearly incidents and fewer procedural departures.
314.481375 319.601375 A system level variable, perceived procedure quality, was overall the best predictor for
319.601375 322.161375 all outcome and behaviour variables.
322.161375 327.761375 Predictably, higher industry experience was related to less procedure use and greater
327.761375 329.161375 rule departures.
329.161375 333.401375 Perception of higher procedure quality was related to fewer incidents and near misses
333.401375 336.281375 in both individual and system models.
336.281375 343.161375 So generally, and maybe not too unexpectedly, the more positive people saw the procedures,
343.161375 345.881375 the more likely they were to actually use them.
345.881375 350.681375 The findings also highlight the importance of procedure quality, as it’s the best and
350.681375 355.761375 unique predictor of both procedural deviation and procedure use behaviours.
355.761375 360.321375 So these findings indicate that not only do system level variables predict procedure related
360.321375 365.481375 behaviours and outcomes, like injuries and incidents, but they uniquely predict above
365.481375 370.641375 and beyond individual level characteristics, suggesting the importance of system variables.
370.641375 377.001375 Further, to this point, is that a model one view of procedures may be necessary, but insufficient
377.001375 381.361375 for understanding and predicting procedural system effectiveness.
381.361375 386.401375 Said differently, a view of procedures as strict prescriptions of the idolised way of
386.401375 391.001375 working isn’t exclusively valid based on these results.
391.001375 395.281375 When workers perceive procedures as tools to support them, rather than as prescriptions
395.281375 399.761375 to follow or control their behaviour, then they’re more likely to use them, more likely
399.761375 404.201375 to adhere to the procedures and the outcomes are more favourable.
404.201375 409.001375 Other system variables like safety climate and the change process evaluation, you know,
409.001375 413.361375 change management around procedural changes, didn’t uniquely predict any behaviours or
413.361375 415.841375 outcomes that were studied in this paper.
415.841375 421.081375 A further interest is that, while reported procedural deviations are correlated to the
421.081375 426.641375 incidents, they don’t provide more predictive ability beyond perceived procedure quality.
426.641375 432.441375 That is, procedural departures are often due to the procedures quality rather than any
432.441375 438.161375 other issues related to people, they just don’t care enough, they’re lazy, they complacent.
438.161375 442.321375 These things didn’t predict the procedure years or the procedural departures, it was
442.321375 446.621375 the perception of the quality of the procedures that did.
446.621375 448.601375 So what can we make of the findings?
448.601375 453.461375 Reports on this study and several other studies from the same research group suggest that
453.461375 459.281375 there’s value in considering rules not as concrete inflexible constructs, but as knowledge
459.281375 462.041375 reservoirs or resources for action.
462.041375 467.481375 They should help sensitise people to how the work needs to be performed or should be performed
467.481375 473.401375 or could be performed, what to do, how to do it, when to do it, particular things that
473.401375 476.001375 must be done or must not be done.
476.001375 481.601375 If a rule is important, as in, the work has to be done in a particular way or sequence,
481.601375 487.401375 then make it known, ensure people are set up to succeed with the task and the procedure.
487.401375 493.061375 On that point, there’s really good guidance out there on how to construct effective procedures.
493.061375 497.281375 And here’s a rule of thumb, if these sort of guidelines haven’t been used to construct
497.281375 503.241375 your procedures, then there’s probably a decent chance that your procedures suck.
503.241375 508.041375 Good idea, not discussed in this paper, is breaking up your rules into different types.
508.041375 512.801375 This is discussed again by Hale and Boroughs in other work, but they differentiate between
512.801375 518.161375 performance goals, or performance rules, process rules and action rules.
518.161375 522.881375 Performance-based rules define only what has to be achieved and not how it must be done.
522.881375 528.241375 So this could include just general duties or so far as reasonably practicable.
528.241375 533.381375 Performance rules define the process which people have to follow, but leave considerable
533.381375 536.761375 freedom about how they go about it.
536.761375 543.601375 For instance, the requirements to prepare a risk assessment, or to set up an SMS, or
543.601375 548.401375 requirements to consult with certain people when an emergency situation arises.
548.401375 550.601375 Third, action rules.
550.601375 555.241375 These tend to be the most prescriptive, and they’re usually specified in kind of if-and-then
555.241375 556.241375 statements.
556.241375 560.561375 They really prescribe in detail exactly how people should behave, or how the hardware
560.561375 562.921375 should be designed or tested.
562.921375 565.801375 Action rules provide the least flexibility.
565.801375 570.401375 They really should only be used sparingly for things that are absolutely generally critical,
570.401375 574.401375 or need to be done in a very specific way in a particular sequence.
574.401375 579.641375 Unfortunately, a lot of the rules that we default to in safety tend to be action rules, when
579.641375 584.121375 what we really should be defaulting to are more performance or process rules.
584.121375 588.641375 Also, consider the use of meta-rules, again, not discussed in this paper.
588.641375 595.301375 But meta-rules are rules about how to use rules, or more specifically for this context.
595.301375 602.281375 When can we work around the rule, like dispensations for rules, or modifications to rules?
602.281375 605.441375 Simple heuristics and rules of thumb can work really well here.
605.441375 610.321375 Dave Snowden, famous from the Canevan, has some really interesting ideas on meta-rules
610.321375 616.101375 and heuristics around procedural departures and procedural dispensations.
616.101375 622.701375 But as a final caution, be careful with confusing prescribed systems with reality.
622.701375 628.961375 Procedures are only ever incomplete abstractions of work, they’re not the work itself.
628.961375 634.121375 Paraphrasing the authors Pentland and Feldman, it’s a folly of designing artefacts while
634.121375 640.201375 hoping to change patterns of action.
640.201375 645.281375 Or, as Dadirio said, documents aren’t the routine, that is, “No matter how beautiful
645.281375 650.241375 they are, no matter how much you work on them, those procedures are never the things that
650.241375 651.241375 people do.”
651.241375 658.081375 So overall, this study suggests that Model 1 and Model 2 views of procedures are complementary
658.081375 661.041375 and not conflicting, they both have their role.
661.041375 665.961375 Further, only focusing on procedural quality in itself, you know, as in trying to create
665.961375 670.821375 the perfect procedure, without accounting for the utility of the procedure as experienced
670.821375 676.361375 by workers, may fail to make any significant progress, and thus highlights the need for
676.361375 681.721375 workers to create useful procedures as defined by themselves.
681.721375 686.841375 So limitations, as you’d expect, interviews and cross-sectional surveys, they can really
686.841375 688.441375 only get you so far.
688.441375 693.601375 The authors note that it would have been useful to have some observational data and probably
693.601375 698.041375 a use of some naturalistic or ethnographic studies of field work and use of procedures
698.041375 699.121375 would be really useful.
SafetyInsights.org 