When safety interventions backfire: Why some measures can have very unexpected results

The net impact of interventions that prevent or reduce the associated costs of risky behaviors are not always positive. In some cases, protection measures can lead people to „forget to be afraid“, resulting in higher risk taking and negative net results. In the current research, we focus on safety measures (such as safety helmets and seatbelt legislation) to clarify the conditions that drive this phenomenon. Using a controlled laboratory experiment, we found that protecting people from a rare disastrous event was effective and did not backfire. Conversely, preventing frequent moderate losses from risk taking induced higher risk taking rates, impairing participants’ earnings. We discuss the implications for terror networks and organized crime.

The current study was designed to clarify the conditions under which safety enhancing interventions are likely to backfire. The research was motivated by a recent increase in road fatality rates in the USA – although road death rates were in a steady decline for decades, since 2011 this trend has changed and death rates have been slowly climbing (NHTSA, 2017).

One natural explanation for the increase in fatal road accidents highlight the negative side effects of new technologies, such as smartphones, that increase the benefits of risky behaviours.  For example, it is possible that in certain cases texting while driving maximizes the driver’s expected utility even when it increases the probability of an accident. This „rational cost-benefit explanation“ suggests an easy solution to the apparent problem: Reducing the benefit of new reckless behaviors with improved law enforcements and bigger punishments.

The current paper examines the feasibility of a second contributor to the disturbing road fatality pattern. It considers the possibility that part of the problem reflects the impact of new technologies that were designed to enhance safety, but backfire. Specifically, the current paper analyses situations where the decision maker is partially protected from harmful events that an undesirable behavior[1] entails.

In our analysis, we distinguish between two classes of partial protection. The first involves partial protection that improves the worst-case outcomes. Examples (for car drivers) are safety helmets and airbags. For a Terrorist Networks (TN) and Organized crime (OC) related example, imagine the effect of abolishing capital punishment (a relatively rare procedure) on a country’s murder rates.

The second class involves partial protection from mild but frequent losses.  Examples of technology that provide such partial protection (implying a typically forgiving setting) include shock absorbers and Anti-lock Braking systems (ABS) in cars. From a TN/OC perspective, one example are situations when a country forgoes or is unable to punish minor criminal infringements with high probability, while inflicting very high punishments on rare cases. This can facilitate and encourage criminal activity.

Basic research in psychology and behavioural economics suggests that both subclasses of partial protections can backfire.  Improving the worst-case outcomes is likely to backfire when behaviour is driven by the Peak-end rule (Fredrickson and Kahneman, 1993).  The Peak-end rule suggests a tendency to remember the peak (extreme) and the end (final) experiences while underweighting other experiences. For example, if people avoid participating in TN/OC activities because of the threat of capital punishment, eliminating the risk of a death sentence will encourage TN/OC involvement.

Preventing mild and frequent losses from risky choice is likely to backfire when people exhibit underweighting of rare events (see Barron and Erev, 2003; Etzioni et al., 2017; Hertwig and Erev, 2009). A bias toward underweighting of rare events tends to emerge with accumulation of experience. If people avoid TN/OC activities because it results in social exclusion that implies frequent small costs, reducing the costs of social exclusion might encourage TN/OC involvement.

Key findings

Our study shows that giving people protection against rare, large losses is highly effective (i.e. does not induce higher risk taking), while protection against common, moderate losses can drive them into taking counterproductive risks that impairs their expected returns. In the paper, we also shed light on the relative importance of two behavioural tendencies (overweighting the peak experience, and underweighting of rare event), and on the value of a simple model that assumes people rely on small samples of experiences to make decisions.

Negative net effects from partial protection are particularly likely when the intervention eliminates small, common losses.  In our experiment, the elimination of small frequent losses that improved the expected value from risky choice by ~30% reduced participant’s net results by more than 100%.  In contrast, an intervention that improved the worst possible outcome (we manipulated this intervention to have same effect on the expected value as the previous intervention) increased the participants’ average payoff by 20%.

We believe that the higher descriptive value of the underweighting of rare events hypothesis over the peak-end rule reflects the fact that in the current context, implicit memory is more important than explicit memory.  Previous research shows the peak-end rule is highly descriptive in situations where the subjects are explicitly asked to rank their experiences, but the tendency to underweight of rare events is more likely to derive the implicit impact of memory on ongoing decisions (Schurr, Rodensky and Erev, 2014).  Thus, the tendency to “forget to be afraid” appears to be a reflection of implicit memory.

Relationship to Terror Networks and Organized Crime

These results suggest many implications for stakeholders dealing with TN/OC. For example, it might facilitate a better ability to foresee potential effects of interventions and measures on the motivations of TN/OC operatives. For example, in some situations, conceding worst-case punishment can have minor consequences on the motivation to commit crimes, while improving greatly law-enforcement relationship with the community. Improving the chances of incurring frequent losses and costs for undesirable behavior might reduce individual’s motivation to partake in such activities.

Our results can be further used to facilitate the design of more effective safety interventions for the public. For interventions meant to protect people from the possible results of their behaviour, effectiveness is expected to increase by focusing on the worst-case scenarios. For interventions meant to prevent undesired behaviours, focus on frequent costs is recommended.

The current results also suggest that consumers might prefer frequently-forgiving technologies to the technology that improves the worst-case outcome (or even stop using technology the improves the worst-case, even though they say they plan to use it; see Yechiam, Erev and Barron, 2006). To reduce the risk such bias entails, it is probably necessary to monitor the required behavior and reduce or improve its apparent attractiveness, using the framework outlined above.

Note: This article is based on the paper “On Safety, Protection, and Underweighting of Rare Events” by D. Cohen & I. Erev (2018), currently under review in Safety Science.

Author

Doron Cohen, Ido Erev (TECHNION, Israel)

References

Barron, G. and Erev, I., 2003. Small feedback‐based decisions and their limited correspondence to description based decisions. Journal of Behavioral Decision Making, 16(3), pp.215-233.

Etzioni, S., Erev, I., Ishaq, R., Elias, W. and Shiftan, Y., 2017. Self-monitoring of driving speed. Accident Analysis & Prevention, 106, pp.76-81.

Fatality Analysis Reporting System (FARS). (2017). National Center for Statistics and Analysis (NHTSA). Home | National Highway Traffic Safety Administration (NHTSA) | U.S. Department of Transportation, retrieved from: https://www-fars.nhtsa.dot.gov/Main/index.aspx

Hertwig, R. and Erev, I., 2009. The description–experience gap in risky choice. Trends in cognitive sciences, 13(12), pp.517-523.

Kahneman, D., Fredrickson, B.L., Schreiber, C.A. and Redelmeier, D.A., 1993. When more pain is preferred to less: Adding a better end. Psychological science, 4(6), pp.401-405.

Schurr, A., Rodensky, D. and Erev, I., 2014. The effect of unpleasant experiences on evaluation and behavior. Journal of Economic Behavior & Organization, 106, pp.1-9.

Yechiam, E., Erev, I. and Barron, G., 2006. The effect of experience on using a safety device. Safety science, 44(6), pp.515-522.

[1] behavior such as texting while driving and (under some assumptions) joining criminal and terrorist organizations.