Cognitive biases apply to tailings discussions
Nov 27th, 2019
Cognitive biases apply to tailings discussions more than one could think. Indeed, cognitive biases can cloud debates and oftentimes lead to worrisome “alternative facts” bordering with fake news.
Cognitive biases are likely hidden in glossy documents, websites, seemingly authoritative statistical approaches which are then oftentimes referenced by inattentive readers. The final effect is of course to perpetrate the cognitive drifts.
Original authors foster the noble cause of reducing the risks posed by the hazardous structures used to contain tailings, i.e. tailings dams. While we certainly share that cause objective, and we have devoted huge efforts to it, we also think tailings discussions should remain “real” and not drift toward alternate reality.
Cognitive biases apply to tailings discussions
Mixing apples and bananas
We read in various documents that the number of tailings dams is way larger than the “good old” 3,500 active dams, which by the way, we used back in 2013 at the tailings and mine waste conference. Incidentally, the Church of England inquiry approximatively reconfirmed that number. However, the number of 18,000 to 20,000 is popping up here and there. It is very likely that humankind has indeed created so many tailings dams during the Anthropocene, once we stopped throwing everything into rivers.
The problem is that such a large inventory likely includes more than one population of dams “size, volume, potential impact”. If we accept this mixing “apples and bananas” the likelihood of catastrophic failure per year per dam would go down by one order of magnitude. If that was true, then experience says public opinion, media, politicians would not raise their arms like they do now.
- The bias pushes authors to inflate the number of dams to show the problem is important, but the final result is actually counter-productive for the cause!
- To really promote safety and risk reduction we need to eliminate cognitive drifts.
Another example? We see “statistical approaches” mixing catastrophic releases of various types, disregarding the nature of the accidents. For example, the Samarco (Chapter 12 of Tailings Dam Management for the Twenty-First Century) tailings release entered a river. Thus one cannot mix the large distance covered by part of the tailings with other failures on “dry land”. This mixing generates confusion and clouds the issues.
New codes and documents propose what their authors believe are prudent evaluations of consequences. They think they are erring on the side of safety, which is a good intention.
- The problem is the following: most of these documents propose various categories (dimensions) of consequences, like, for example human, environmental, infrastructural, etc. They advise the user to qualify the dam consequence using the highest one.
- We will not discuss here the fact that the “scales” of these categories oftentimes present logical and quantitative inconsistencies.
- Let’s also note that these codes only propose the low-bound of the casualties of the “extreme” category. Thus, these codes put in the same category dams that could generate say 1,000 casualties and 100,000 casualties. That cannot be right!
Think about this. A catastrophic failure includes all the consequence dimensions, so how can we only use the worse? Ignore that the sum of “high” dimensions may by far exceed the “extreme”? This is a worrisome bias indeed.
Using only consequences to prioritize dams portfolios
If we applied new tailings dams articles and proposed codes to hydro-dams in the Alps, all should be fixed immediately because their consequences are “extreme”! We are not going to discuss here the fact that in some countries those dams may be poorly maintained and aging. That’s a different subject.
Why are public opinion, politicians, media not raising against hydro-dams in the Alps? Well, simply because their rate of failure is minimal, nearing the credibility threshold.
What does that mean? Simply that although the potential consequences of their failure are undoubtedly in general catastrophic, their probability of failure is deemed very low, bordering credibility.
- Using consequences only to prioritize dams portfolios biases prioritization!
- If we seek to mitigate harm to people at planetary scale while remaining rational, we have to delete the “consequence only” clause in new codes.
Not using quantitative risk
This point is almost a result of the prior ones. Quantitative risk is most times avoided in mainstream practice. Yet we have proven many times that obtaining probabilities estimates is not a daunting task and uncertainties can be included.
Suppose you have under your responsibility a very modern, well investigated, designed, built and monitored tailings dam. The estimate probability of failure is less than 10-4 per year. Population at Risk (PAR following modern codes) is 100, thus the dam is “extreme” despite the good estimated reliability.
Risk (human harm only) is 10-4*102=0.01
Next dam in the inventory impinges on the same valley, hence has also PAR=100.
However, it is an older dam, with poor probability of failure, say 0.5*10-3.
The risk (human harm only) is 0.5*10-3*102=0.05.
That is this second dam has 5 times more risk than the prior: yet new codes would not give any rational tool to decide on which one to act first and would triage them equally!
Furthermore, codes keep referring to Factor of Safety (FoS) which is not a measure of reliability. Then they ask engineers to reduce probability to the lowest possible level, but offer no clues on how to do it.
Closing remarks on Cognitive biases apply to tailings discussions
As we recognize that Cognitive biases apply to tailings discussions there are no excuses to perpetrating them.
A prominent expert in tailings dams I have in high respect said during a round table at TMW2019, it is time we avoid misleading codes! If I am not citing his name is simply because I did not have time to ask permission. However, I am sure he will recognize himself if he reads this post.
Thanks for your wisdom, M!
Tagged with: cognitive biases, tailings, Tailings Dam failure
Category: Hazard, Mitigations, ORE2_Tailings, Probabilities, Risk analysis, Risk management