Latest / Safe As Podcast / E68: Do warning signs work? A dive into the science

Transcript

E68: Do warning signs work? A dive into the science · Safe As Podcast

• 0:00
  You saw the warning sign and read the text, and apparently
• 0:04
  you know the risk. So why did you do it anyway?
• 0:07
  It turns out there's a notable gap between knowing a danger
• 0:10
  exists and actually changing your behaviour to avoid it.
• 0:13
  So what makes an effective warning?
• 0:16
  Good day everyone. Welcome to Safe as I'm Doctor
• 0:19
  Ben Hutchinson. Today, we're diving into a meta
• 0:22
  analysis that explores 30 studies to determine how
• 0:27
  effective are warnings. A meta analysis is a technique
• 0:30
  where the results from multiple studies on the same subject are
• 0:35
  grouped together to evaluate overall trends.
• 0:38
  The reference is Hancock ET al, 2020, titled How effective are
• 0:44
  Warnings? A Meta Analysis published in
• 0:47
  Safety Science. Warnings include multiple
• 0:50
  things, including hazard labels, science, safety, data sheets,
• 0:56
  medication and more. Warnings contain explicit
• 0:59
  information that query in the paper communicates the existence
• 1:02
  of a threat of harm. Warnings are part of the
• 1:05
  hierarchy of control. So while physical barriers can
• 1:08
  be seen as the first line of defence, warnings are towards
• 1:12
  maybe the last lines of defence and unfortunately they persist
• 1:17
  as a relatively weak form of protection.
• 1:20
  The authors separate warnings from other forms of hazard
• 1:23
  communications. Warnings are different from
• 1:26
  advisories, which communicate anticipated, preferred, or
• 1:30
  expected courses of action. Alerts are also different, which
• 1:34
  convey useful information but not the specific information
• 1:37
  related to the anticipated harm. So warnings are more targeted
• 1:41
  and specific than alerts, though the author's note there is some
• 1:44
  overlap between these concepts. Finally, they say that while
• 1:48
  physical harm seems to dominate in warnings, there are other
• 1:52
  forms like viewer discretion is advised.
• 1:55
  This is an example of a warning against cognitive harm.
• 1:59
  Before we get into the specific findings of this study, let's
• 2:02
  cover some of the key variables potentially influencing the
• 2:06
  effectiveness of warnings. The meta analysis included these
• 2:09
  following factors, integration of the warning into the task,
• 2:13
  it's seriousness or specificity of consequences like a warning
• 2:18
  starting with serious injury or death, the warning visibility
• 2:22
  and if it's eye catching, details in the warning and
• 2:26
  clutter in the writing and more. So what did they find?
• 2:30
  A key finding of course, is that there's a gap between knowing
• 2:35
  and doing. So we can compare recall versus
• 2:37
  compliance. The study confirms a significant
• 2:40
  gap. Recognising a warning does not
• 2:43
  linearly equate to complying with it.
• 2:46
  Someone can perfectly recall a warning's content but still
• 2:50
  choose to ignore it. A warning's noticeability is
• 2:53
  mediated by its size and sensory noticeability.
• 2:57
  Things like its luminance, bright colours and contrasting
• 3:00
  backgrounds. Recall of content is
• 3:03
  significantly improved by these types of eye catching sensory
• 3:08
  features quickly. Based on the forest plot in the
• 3:11
  study, a forest plot's used to visualise results.
• 3:15
  The findings further to the right of the plot indicate
• 3:18
  stronger positive effects on compliance.
• 3:21
  The presence of words on warnings shows the largest
• 3:24
  estimated effects, with task integration and warning
• 3:28
  placement also demonstrating significant positive impacts.
• 3:33
  The horizontal lines under the numbers or whiskers represent
• 3:37
  the uncertainty around each effect estimate, so there's some
• 3:42
  imprecision in the results. Importantly, when those whiskers
• 3:46
  cross the centre line, we can't conclude that the effect is
• 3:49
  statistically significant. It's also worth noting that many
• 3:53
  of these confidence intervals overlap as horizontal lines,
• 3:57
  which means we should be cautious about ranking warning
• 3:59
  features as clearly better or worse because we just don't have
• 4:02
  strong enough evidence. So what drives adherence to
• 4:07
  warnings? First, the power of words.
• 4:10
  Interestingly, the presence of text is a much stronger
• 4:13
  predictor of compliance than pictures alone.
• 4:16
  Explanatory text provides an immediacy of understandability
• 4:20
  that graphics often lack. Second, task integration.
• 4:25
  Compliance is highest when the warning is unavoidably
• 4:28
  encountered during a task. For example, placing a warning
• 4:32
  label directly on the tape used to open a package is more
• 4:36
  effective than placing it in a separate manual.
• 4:38
  Third, proximity and immediacy warnings work best when they
• 4:43
  access direct perception at the moment of risk, rather than
• 4:46
  relying on a user to recall a warning that they saw earlier.
• 4:50
  Next is about the psychology of warnings.
• 4:53
  Humans are information foraging creatures who quickly stop
• 4:57
  paying attention to uninformative sources.
• 5:00
  If a warning is omnipresent, like it's always around us but
• 5:04
  never leads to a consequence, people habituate to it and stop
• 5:07
  noticing it entirely. Next, people rate warnings as
• 5:11
  more believable and treat them more seriously when the language
• 5:15
  is more specific about the hazard.
• 5:17
  Phrases like this product will kill you, remove ambiguity, and
• 5:21
  increase the perceived threat. Finally, there's an anti warning
• 5:26
  paradox, meaning there's a conflict between safety and
• 5:30
  sales. A warnings present can decrease
• 5:33
  a person's positive attitude towards a product.
• 5:36
  This leads to anti warnings cases where commercial interests
• 5:39
  may intentionally weaken a warning to avoid discouraging a
• 5:43
  purchase. So what can we make of the
• 5:45
  findings? For one, there's an art and
• 5:49
  science to designing effective warnings.
• 5:51
  Luckily, there are plenty of guides available.
• 5:54
  Drop a comment if you want me to cover one of those guides in a
• 5:57
  future episode. Secondly, the traditional
• 6:00
  hierarchy of control lists warnings among the least
• 6:03
  effective options and really the last or second last line of
• 6:07
  defence. But interestingly, these authors
• 6:10
  argue that regarding psychological harm, warnings can
• 6:14
  be considered as a first line of defence.
• 6:17
  I'm not so sure I agree, given the research highlighting how
• 6:22
  design organisational and structural factors within
• 6:26
  organisations can create the conditions for psychological
• 6:30
  harm. But it's an interesting
• 6:31
  contention anyways. To optimise warnings, integrate
• 6:36
  them into the task as best you can.
• 6:38
  Use clear text over pictures and be specific about the
• 6:42
  consequences and the location matters.
• 6:45
  Placement has a strong effect. The study notes that horizontal
• 6:49
  warnings are generally recognised faster than vertical
• 6:53
  ones, and placement to the left of centre can lead to faster
• 6:57
  response times. Clutter is also a compliance
• 6:59
  killer. An increase in clutter like
• 7:02
  crowded labels or small fonts significantly decreases
• 7:06
  compliance. Simpler warnings are not only
• 7:08
  complied with more often but are also recalled with higher
• 7:12
  accuracy. And shameless plug but semi
• 7:16
  related, check out my audio pod on Spotify or Apple for research
• 7:21
  on designing better procedures. It's self and art and science.
• 7:26
  That's it on Safe As. Please like and subscribe and if
• 7:30
  you'd like to support the channel then feel free to share
• 7:32
  a coffee. Links in the show notes.