By Laura McGuire and Jesse Percival
This article originally appeared in The Avalanche Journal, Volume 119, Winter 2018-19
COGNITIVE SYSTEMS ENGINEERING (CSE) demonstrates how expert practitioners in high risk/high consequence domains make sense of risk in dynamic, ambiguous and changing conditions. Expert performance is identified as going beyond qualifications to include the ability to activate, organize and flexibly apply knowledge (Woods et al, 2010) in time pressured, goal conflicted and uncertain conditions. To do so involves cognitive work.
Using methods from CSE, this study assessed the operational aspects of snow safety then analyzed the artifacts (tools such as worksheets, websites, whiteboards, InfoEx, etc.) that shape cognition and collaboration. Semi-structured interviews were used to detail how tools are used to make and update forecasts over time. Finally, we elicited examples of surprise, near misses and actual incidents to calibrate findings.
Three prominent, interconnected themes emerged from the research:
- Much of the cognitive work is not described in the explicit protocols. The formal representations of what constitutes good practice in forecasting is a small fraction of the strategies experts use.
- The cognitive effort required to manage avalanche risk is a near continuous activity. Forecasting appears to require ongoing calibration. Disruptions to this calibration process have adverse effects on performance.
- Forecasting is a distributed cognitive task across individuals, teams and the broader industry. Successful forecasting requires distributed practitioners of local team members as well as the resources and insights produced by others within the industry.
PREPARATIONS FOR FORECASTING
Formally, the protocols for a forecaster on duty (FOD) suggests producing a control plan shortly after arriving onsite – but each forecaster interviewed detailed extensive preparations that were not captured by the formal description. A variety of work-related techniques were described. For example, time spent carpooling is used as an informal handoff from one FOD to another to discuss recent activity or control measures. This suggests that formulating the day’s forecast begins well in advance so that a forecaster arrives for duty with a hypothesis of how recent changes in conditions affect their avalanche terrain management.
Shared, off the books activity is a common (and likely necessary) practice not explicitly noted in work procedures and demonstrates a need for ongoing calibration – an example that supports all three findings. It is well documented that forecasting takes place under time pressure. By seeking out data that can help them anticipate conditions in advance, the FOD relieves some of this pressure to lessen the cognitive demands required once they officially clock in.
DISRUPTION, ADAPTATION & SURPRISE
A second example: An unexpected in-bounds release. On this day, the forecasting plan had anticipated instabilities due to temperature changes. After control work, it was expected that normal monitoring would identify if a closure was necessary. However, a personal emergency meant the team was operating one person short. Concurrently, a first aid emergency tied up members who would otherwise be monitoring avalanche terrain. This left the FOD ‘in the bump’ for longer than the usual rotation and his normal practice was interrupted. As expected, the temperature fluctuated and a skier-triggered release occurred in one of the avalanche zones.
This example is informative in two ways. Firstly, it is reflective of what “normal work” is – constantly adjusting to workload demands or unavailability of resources and adapting practices to respond to conditions while balancing inevitable tradeoffs. Secondly, this example provides evidence that practitioners construct mental models (Adams, 2005) and continually update them.
MENTAL MODELS
The model is an internal representation of current hazards and an expectation of how this may change over time. Mental models are used to retrieve technical knowledge and to flexibly apply it to variable situations.
In constantly changing conditions, mental models become stale unless continually updated. Referring to the in-bounds avalanche example, the model became insufficient after only a few hours. In the previous example, the forecaster coming back from time off is aware their model is stale and seeks information to recalibrate. LaChapelle (1980) notes a “…prevalent and strong reluctance of working forecasters to experience an interruption in their winter routine…” (pg. 78). This finding emphasizes organizing work schedules to protect forecasters’ daily and seasonal monitoring routine from interruptions or building in mechanisms to support rapid recalibration or redundancy by cross-checking across other team members.
DISTRIBUTED COGNITIVE EFFORTS
Notable as well, is the role of a distributed network in constructing mental models. A diverse range of perspectives informed by different experiences, knowledge and mindsets is needed for accuracy. In the resort, the schedule for FOD’s is designed to provide an overlap day to accommodate the need for distributed cognition. This is an explicit recognition of both ensuring currency of the mental model and the importance of interactions between practitioners. Updating provides an opportunity to draw attention to details and to generate shared insights.
Spatial and temporal constraints also require distributed cognitive efforts. Large terrain and limited daylight hours create time pressures. The FOD relies on technicians to gather and relay data efficiently and accurately. Without the team, the FOD’s mental model can only partially represent actual conditions.
CONCLUSIONS
Errors by normally high performing experts are insights into how the cognitive demands may become temporarily overwhelming. Studies like this illustrate what aspects of practice should be protected from the pressures of ‘faster, better, cheaper’ common in many workplaces and allows for better engineering of the tools, technologies and protocols used.
Further research can provide an empirical basis for: designing decision support tools; developing training; orchestration & distribution of tasks; funding critical resources; and developing new forms of coordination across networks. Identifying cognitive work in different forecasting settings (mechanized skiing, transportation, industrial) is likely to be useful for accident prevention. In addition, CSE studies comparing expert vs recreational cognition is likely to help public safety efforts.
The authors gratefully acknowledge the Avalanche Canada Foundation for their travel support through the ISSW Fund and the Cora Shea Memorial Fund. For the complete proceedings paper or more information about this and other projects in cognitive work of avalanche forecasting contact Laura or Jesse.
REFERENCES
Adams, L. (2005). A systems approach to human factors and expert decision-making within Canadian Avalanche Phenomena. MALT Thesis. Royal Roads University, Victoria, BC, 284.
LaChapelle, E. R. (1980). The fundamental processes in conventional avalanche forecasting. Journal Glaciology, 26(94), 75–84.
Woods, D., Dekker, S., Cook, R., Johannesen, L., Sarter, N. (2010). Behind Human Error. London: CRC Press.
