Skip to content

Trust in Allah, But Tie Your Camel, by Russ Costa

by Russ Costa      

This is a continuation from the Blue Ice avalanche from January 2019. You can find the accident report and my introduction “Low Danger – Chapter 1” in the essays.

Having spent the morning of Saturday, January 5th sitting in a classroom, I was excited to get up in the mountains and onto the snow in the afternoon. I was observing a backcountry clinic that day, collecting some data, and thinking about how best to teach decision-making in avalanche terrain. Although there wouldn’t be much skiing today, it would be good just to stretch my legs and breathe mountain air. I had read the Salt Lake area avalanche forecast in the morning over coffee, as is habit. The avalanche danger was low across all aspects and elevations. A “green light” day, as we call it. Not that it mattered much for me today – we wouldn’t be venturing into significant avalanche terrain. I looked out the window at the mountains and my mind wandered to what I could be skiing if I were not observing this clinic.

The afternoon was colder and windier than I had expected. It wasn’t inhospitable, but it wasn’t the ideal weather in which to be standing around in the mountains either. Thankfully, I had a down coat stuffed away in my backpack. We gathered just above treeline, but well below the ridgeline. There, we were at least relatively sheltered from the wind, which appeared to be gusting on the ridges and peaks. I was glad we weren’t up there, but not for safety reasons so much as for comfort ones. The snow was unexpectedly grabby, as a thin wind skin had formed over its surface. Small cracks broke out as we skied and skinned across it. None of us were particularly concerned though, as the affected top layer wasn’t more than a few inches deep, and the terrain we were on wasn’t steep. Plus, it was a “green light” day. Nonetheless, I silently looked at the terrain above where our group had gathered, happy the slope above us was windward and not loading. It was a low danger day, but sometimes I look left and right when I proceed through a green light.

The next morning, I read the forecast with the reports from the 5th of January and, it being 2019, the social media response to it. Eight skier-triggered avalanches. Four persons caught and carried in separate events. Only one injury and no fatalities, thankfully. Reminders that aren’t full-fledged disasters are probably a good thing for a community that (at the time) hadn’t experienced an avalanche-related death in almost three years, despite seeing tremendous growth in the popularity of the sport. But what happened that day? With the forecast? With the weather and the snowpack? With the human factors? I’ve been trying to understand all of those questions – especially the latter one.

When it comes to judging monetary values, from real estate values to civil litigation damage suits, research in cognitive science and behavioral economics has demonstrated that humans consistently anchor their judgments of value closely to some preset number, and then subsequently adjust their estimates of the appropriate value too conservatively (their estimates remain too closely tied to the anchor value) in light of new information. I think a similar cognitive process operates when we estimate danger hazard and risks in the backcountry, and that forecasted danger levels often serve as such an anchor value.

The anchoring and adjustment heuristic operates when decision-makers are compelled to make estimates (of value, of frequency, of danger, etc.) under uncertainty. In a classroom demonstration, I ask students to estimate the number of cells in their brain, the number of shark attacks per year, and other such values that they are likely highly uncertain about. They are consistently and significantly swayed by estimates (or “averages”) I provide them with, although I just made those estimates up for the purposes of the demonstration. If I tell students that there are an average of 1,000 shark attacks per year worldwide, they will estimate that there were between 900 and 1,100 last year; however, if I tell them there are an average of 10 shark attacks per year, their estimates typically fall between 8 and 12. They adjust their values slightly based on new information (i.e., news coverage or lack thereof of shark attacks), but their estimates remain very closely tied to the “anchor” values that I essentially planted in their minds.

We are always uncertain about actual risk in the backcountry and as such, must estimate it based on prior knowledge and observations.  Specifically, in judgments of avalanche risk, the forecasted danger rating (low, moderate, considerable, high, etc.) at the relevant aspect and elevation can serve as an anchor. If the forecasted danger rating is “low,” for example, backcountry travelers may be less likely to adjust their evaluation of danger significantly upward toward “moderate” or “considerable,” even upon encountering widespread information in their travel indicating that the danger may be higher than “low.” In this case, their judgment of risk, and their mental model of hazard remains too tightly anchored to the forecasted danger rating.

Anchoring acts in conjunction with other cognitive biases and heuristics. It is likely that this anchoring effect is asymmetric in backcountry decision-making. Backcountry skiers and riders seem less likely to adjust their evaluation of danger upward (toward more hazard) from the forecast anchor than they are to adjust their estimates of risk downward, toward a lower risk judgment. This behavior can be explained by what Ian McCammon (2002) calls the commitment (or consistency) heuristic. Adjusting risk estimates toward higher danger would be generally inconsistent with objectives or goals the individual or party has committed to for the day, and is thus less likely to occur, while adjustments toward lower danger would be consistent with the goals or objectives for the day.

We not only anchor our estimates too closely to the forecasted danger, but we are also more likely to notice information that confirms that hypothesis or estimate. If we expect the danger to be low, or it was forecasted as low, we are likely to notice signs of stability or, more importantly, ignore (or not notice) signs of instability. This tendency to search for, interpret, and remember information that confirms one’s preexisting beliefs, hypotheses, or wishes (and to ignore or minimize information that disconfirms them) is known as the confirmation bias (Wason, 1960). In past talks for the avalanche community, I’ve described how the confirmation bias can influence our snowpit observations: we see in the snow what we believe or want to be true. I think the confirmation bias can also operate at the level of the avalanche forecast: we go out looking for information to confirm what we read in the forecast, especially if that forecasted danger is low and is consistent with our goals and objectives for a particular ski tour.   

So, what can we learn from this day? Below are three quick human factors lessons, I drew from January 5th, 2019. Decades of research in psychology, cognitive science, and behavioral economics has taught us that humans have many powerful biases. The unifying theme here is not to deny their influence on us, but rather to use rules that lean against them, like a mountaineer sometimes leans into the wind.

1.) Avoid asymmetric adjustment of risk judgments. Like many ski mountaineers, I like to develop ambitious plans in the days or weeks leading up to a tour. I often find myself having to scale back those plans or objectives once I’m “on the ground,” opting for a more conservative descent (or even ascent) routes. This is part of the game. I do not, however, allow myself to scale up plans for the day because the skiing was really good or because the snow seems more stable than I expected. This rule reverses the asymmetry noted above. I can’t say I’ve always followed this rule in my skiing and mountaineering career, but I have obeyed it, more or less, for the last decade, and plan to continue doing so.

2)  Look for disconfirming information. This habit was ingrained in me from my training as a scientist. Good scientists seek not to confirm existing hypotheses or theories, but rather to falsify them. Practice this in your avalanche observations. If the danger is “low,” tune your senses to look for signs of instability on your approach or in your snowpits. I recommend this disconfirming eye should be used to counter hypotheses or beliefs about stability; I caution against using it to disconfirm information that suggests the snowpack might be unstable by searching for clues of safety.  

3) “Trust in Allah, but tie your camel.” (Arabic proverb). Trust the avalanche forecast but put in your own observational and decision-making work in the field to insure your safety too. Don’t rely only on faith in the forecast. We all know weather forecasts can be erroneous. Compare your field observations to the forecasted mountain weather that informed the morning advisory. Especially when more snow accumulates, it is windier, or it warms faster than forecasted, be ready to adjust your judgment of risk.  Your observations on the ground, particularly if you have proper training, are more local in time and place, and thus can be more valid than a generalized report that is indeed, only a forecast.

Further Reading:

McCammon, I. (2002, September). Evidence of heuristic traps in recreational avalanche accidents. In Proceedings International Snow Science Workshop, 244-251.

Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124-1131.

Wason, P. C. (1960). On the failure to eliminate hypotheses in a conceptual task. Quarterly journal of experimental psychology, 12(3), 129-140.

Russ Costa is an Associate Professor of Honors & Neuroscience at Westminster College in Salt Lake City. He studies attention and perception in the lab, and decision-making and risk-taking outside of it, preferably on snow in the mountains.

Leave a Reply

Your email address will not be published. Required fields are marked *