Risk workshops can be tricky events at the best of times. Even if one is successful at herding team members, managers and executives into rooms, it’s a challenge to get the engagement necessary for members to critically evaluate the assumptions underpinning risk and strategic outcomes. When it comes to assessing measures of Likelihood (L) and Consequence (C), the choices often succumb to the psychological phenomenon known as groupthink. This article outlines what groupthink is and how the Governance & Risk Team (GRT) at the Blue Mountains City Council (BMCC) used some simple techniques to mitigate it when conducting risk workshops.
Many people have experienced a situation involving groupthink where, for the sake of harmony, group members strive to reach a consensus opinion. In many cases, members disregard their personal beliefs to agree with the dominant view. Even when members are opposed to the decisions of the group, they tend to keep quiet, preferring to keep the peace out of fear of isolation rather than disrupting the uniformity of the crowd1. Such behaviour results in sub-optimal choices and poor decision-making outcomes.
What causes groupthink? According to Irving Janis, the psychologist that first pioneered the research into groupthink in 19712, there are five factors that, when present, lead to the phenomenon. The group:
- Is highly cohesive with bonds having developed over a long period.
- Is insulated from outside information (or ignores it).
- Is dominated by an assertive leader.
- Experiences stress because of a critical incident.
- Only considers a limited range of options.
What are the outcomes of groupthink? According to Michael W Eysenck3, the above five factors can produce an illusion of invulnerability, in which group members are overly self-assured of their decision-making skill. Members often suppress their opinions to prevent others from expressing ideas, not fitting the collective agreement. For example, before the Challenger disaster in 1986, an engineer who opposed to the launch of the space shuttle was persuaded to change his mind, ‘after being told to take off his engineering hat and put on one representing management’. The results of that fateful decision need no further elaboration.
Members often suppress their opinions to prevent others from expressing ideas, not fitting the collective agreement.
In early 2019, the GRT began implementing the Enterprise Risk Management Framework that had a laborious and, at times, acrimonious gestation. GRT members were new to BMCC following a recent period of upheaval and intense scrutiny of the council. Aware of the situational context and the usual difficulties associated with risk workshop facilitating, the GRT devised a simple experiment to reduce groupthink that would raise the level of engagement, add a dash of fun and improve the chances of obtaining independently derived risk ratings.
The team purchased electronic voting equipment that consisted of remote-control units (RCUs) with buttons representing a Yes/No decision and buttons numbered one to five. The purchase also came with a customisable PowerPoint© presentation that enabled the user to drop in questions that required members to make choices using the RCUs. After the group anonymously voted, the results of the ballot were visually displayed as a histogram in real-time. The PowerPoint© had the inbuilt ability to convert voting choices to numerical averages that represent the independent consensus among the group participants for each decision.
As such, the GRT customised the PowerPoint© to create a single slide for each risk contained in the register. When facilitating a workshop, each group would be required to a) decide whether the risks presented were still relevant to the risk register — a yes/no choice; b) assess the likelihood of that risk, a choice between 1 and 5; and c) evaluate the consequence of that risk, a judgment as per b). Therefore, for any risk, a maximum of three decisions was required. If members voted that a risk was no longer relevant, it was discarded in favour of the next without a vote taking place on Likelihood and Consequence. This would reduce the number of risks to a manageable level, decrease the number of choices required and focus the attention of members on what mattered to them, risk-wise.
Simultaneously, the GRT developed a dynamic Excel©-based 5×5 risk heatmap that allowed for the numerical averages of votes for Likelihood and Consequence for each risk, calculated by the PowerPoint©, to be plotted in real-time. After voting on all of the risks in the register, the group would be able to see where the risk scores (LxC) lie on the heatmap.
The risk workshop
In conducting the risk workshops, the GRT adopted some elements of Janis’ groupthink Framework4. These were:
- Selecting group members whose group cohesiveness encouraged freer thinking
- Splitting the group into independent sub-groups and run simultaneous workshops to assess the council’s risks
- When making decisions about risks (for example, choosing Likelihood and Consequence), members cannot express opinions verbally until after the votes are known
- That members’ role in the workshop is to ‘critically evaluate’ the risks presented and are therefore encouraged to air objections and doubts freely (post voting)
- The Devil’s Advocate position went to members of the GRT who facilitate the risk workshops
The first workshop the GRT facilitated in 2019 was with the Executive Leadership Team (ELT) consisting of the CEO, the directorate heads and the chief safety officer, split into two sub-groups.
The first sub-group consisted of four directors, of which one was a permanent appointee with a long tenure with the council and three others were acting. Two of the acting directors came from previous council managerial positions whose combined tenure was long. The third acting director was an external appointee with less than twelve months’ tenure. As such, this sub-group had a relatively high group cohesiveness.
The second sub-group consisted of the CEO who is new to the position and actively seeks the views of the council’s senior leaders. The second member was another acting director who previously held a managerial position with the council. These two members came from the same directorate, with the acting director being the deputy to the CEO whose combined tenure was long. The third member was the externally appointed chief safety officer with less than twelve months’ tenure. Cohesiveness was, therefore, at a similar level to the first sub-group.
The task before the two ELT sub-groups was to review the enterprise risk register that had a neglected universe of twenty possible risks. Given the decision-making framework outlined above and Janis’ conditions imposed on the workshop’s facilitation, there was a maximum of sixty choices to be made by each sub-group using the voting equipment.
To obtain the best estimates of risk from participants, then groupthink must be eliminated. An effective means of achieving this goal is to let members make their selections by anonymously voting.
During each workshop, members of the sub-groups had to make a decision based on the information contained in the Likelihood and Consequence tables. With the votes tallied, members examined the voting pattern on the PowerPoint©. Where the votes were relatively consistent, as per the Likelihood example is shown in the left-hand panel of Figure 1, the average score became the independent consensus for Likelihood. Where the voting pattern resembled the example in the right-hand panel of Figure 1, then members of the sub-group were asked to discuss their rationale. In most cases, some members possessed information that the others did not, and once this new information was processed, the members were invited to vote again. The results of the second vote then resembled the example in the left-hand pane of Figure 1.
Figure 1: Examples of voting patterns from the workshop
This voting process was repeated for each risk with the scores tallied and plotted on a heatmap at the end of the workshop. The value in each bubble represents the risk number (i.e. Risk #1 to Risk #20). Figure 2 shows the results of the workshop with the first sub-group.
Figure 2: The heatmap from the first workshop
The results of the risk rating exercise show a reasonable spread of results distributed between the Medium and High categories with several outliers and some specific clustering. The average risk score (LxC) was 11.33; the highest-rated risk score was 18 (Risk #19) bordering on Extreme and the lowest rated risk 7 (Risk #20). The aggregated average rating for Likelihood and Consequence were 3.24 and 3.46, respectively, with standard deviations of 0.85 and 0.84. Contrast these results with the ones obtained from the second sub-group, as shown in Figure 3.
Figure 3: The heatmap from the second workshop
There is a marked contrast in location and spread between Figures 2 and 3. Despite the four apparent outliers, the remaining risks are tightly clustered in the Medium and Low categories. The average risk score was 7.82, the highest-rated risk score was 13.33 (Risk #7), and the lowest-rated risk was 7 (Risk #14), indicating very different priorities between the two sub-groups. The aggregated average score for Likelihood and Consequence were 2.44 and 3.13, respectively, with standard deviations of 0.76 and 0.91. The second sub-group believed that the risks presented had a lower probability of occurring with lesser significance to the council compared to the first sub-group.
The GRT then combined the voting data from the two sub-groups and presented a heatmap representing the overall view of the ELT about enterprise risks facing the council. As shown in Figure 4, the assessment of risk is more compact, and there are fewer outliers compared to the independent heatmaps. The average combined risk score was 9.61, the highest-rated risk score was 15.31 (Risk #19 — as per the first workshop), and the lowest-rated risk score was 6.59 (Risk #15). The combined averages for Likelihood and Consequence, and their respective standard deviations lie in the middle of the score obtained from the two workshops as to be expected. Overall, the combined heatmap represents the truest point-in-time, independently obtained estimate of risk at the enterprise level.
Figure 4: The heatmap from the combined votes
One unique element of the voting analysis was the persistence of a weak but positive correlation between the numerical choices selected for Likelihood and Consequence, as shown in the table below.
|Table 1: Correlation between Likelihood and Consequence scores
Theoretically, there should be no association between the numerical choices made for Likelihood and Consequence because each decision is made using different selection criteria. While choices should not be random, they ought to be independently derived.
Given that the choice of Consequence follows directly from the selection made about Likelihood, the persistence of a positive correlation between them indicated that perhaps participants’ numerical choice for Consequence was, in part, influenced by the numerical option selected for Likelihood. As EL Thorndyke postulated, a selection from a given universe of opportunities will, inter alia, increase in probability in proportion to the number of times the same selection was chosen before5. In other words, if I must choose between one and five repetitively, my next choice will be based, at least in part, on my past selections from that universe. Whenever individuals engage in decision-making, they ‘work to reduce the effort they need to expend in making decisions’6. Since members of each sub-group had to make up to sixty decisions, they may have reduced the effort needed on the 40 decisions involving a numerical selection between one and five by selecting the next number with reference, consciously or subconsciously, to a string of past choices.
The GRT examined the distribution of votes in each numerical category for the Likelihood and consequence decision, shown as a combined figure in Figure 5. As can be seen, the choice of ‘3’ out of a universe of numbers between one and five occurred most frequently for both Likelihood and Consequence.
Figure 5: The distribution of votes across contiguous category choices
Was the numerical choice of Consequence influenced by the numerical choice of Likelihood in consecutive decision making? The GRT tested this hypothesis through a regression analysis on the combined voting data which would describe how the Consequence selection is numerically related to the Likelihood selection and indicate the impact of a change in the last choice of Likelihood on the estimated choice of consequence. The regression would also enable the GRT to account for this bias when conducting future risk workshops. The results of the regression did indeed confirm that the choice of Likelihood influenced the successive decision for the Consequence (p-value: 0.0164) in each round of voting on risk. The size of the influence was shown to be present about half of the time or in every second risk evaluated.
Risk workshops can and do work when properly structured. To obtain the best estimates of risk from participants, then groupthink must be eliminated. An effective means of achieving this goal is to let members make their selections by anonymously voting. As shown in this article, the members of two concurrent workshops had very different perceptions of risk, which wouldn’t be present if members succumbed to groupthink. One interesting outcome was that while the results were independent of each member, they weren’t free from individuals’ foibles in decision-making, in this case, the influence of past choices on future choices.