As an experiment, some highly experienced judges were asked to read a description of a woman who had been caught shoplifting. Then they rolled a pair of dice that were loaded so that every roll resulted in either a 3 or a 9. As soon as the dice came to a stop, the judges were asked whether they would sentence the woman to a term of months in prison greater or lesser than the number showing on the dice. Finally, the judges were asked to specify the exact prison sentence that they, using their best judgment, would give the shoplifter. On average, those who had rolled a 9 said they would sentence her to 8 months; and those who rolled a 3 said they would sentence her to 5 months. Somehow, their professional judgment had been influenced by loaded dice.
Seriously? Yes, seriously. This is an example of the anchoring effect, something so common and important in everyday life that you should get familiar with it. It occurs when people consider a particular value for an unknown quantity before estimating that quantity. The estimates tend to stay close to the number considered – hence the term “anchor.” The effect was first widely recognized in a 1974 landmark paper by Amos Tversky and Daniel Kahneman entitled Judgment Under Uncertainty: Heuristics and Biases, and it has been the subject of much research ever since.
Anchoring bias can influence numerous aspects of DoD acquisition programs. In ALE’s own realm of activities, it can appear in reliability predictions, when a reliability analyst is referencing predictions performed on similar systems. It can appear in design review discussions, where a team may latch onto the first alternative identified. It can appear in development (and review) of quotes and proposals. Through ALE’s Quality Management System, ALE staff members are trained to be aware of this bias. The remainder of this article focuses on identifying all the different ways Anchoring Bias can manifest and identifying mitigations so that your teams and companies can overcome it too.
Often, we have to make quantitative judgments without complete information. We are asked to use our professional judgment and experience to arrive at an estimate. One technique is to start with a known number and then make adjustments. If you were asked to estimate what the median price of a three-bedroom home in Central Ohio will be ten years from now, you’d probably want to know what it is today. Then, using that number as a reference point, you’d make adjustments based on assumptions, and apply them to arrive at your estimate. This is a very reasonable approach. But we should be aware of what’s in play when we do that.
Once an anchor is set (i.e., we have some information on a topic) we tend to use the initial reference point as a filter for screening any new information that’s obtained; we assume a range of plausible adjusted values based on this focal point. This can be problematic because it can not only alter our perception, but it can also prevent us from changing our minds in any significant way, even when it may be important to do so. And it manifests itself in several related ways.
One way that anchoring bias leads to error is our propensity to latch onto the first bit of information we get. Consider the various maxims about the importance of first impressions. “You never get a second chance to make a first impression” is an explicit warning regarding the anchoring bias of others. “Don’t judge a book by its cover” is another. All of us tend to give too much weight to first bits of information we encounter, and we latch onto them tightly like security blankets.
A group of high school students was asked to estimate the following product in 5 seconds:
1 X 2 X 3 X 4 X 5 X 6 X 7 X 8 = ?
A similar group was asked the to do the same for this product:
8 X 7 X 6 X 5 X 4 X 3 X 2 X 1 = ?
Of course, the actual answer is the same. However, the average estimate from the first group was significantly lower than the average estimate from the second. Reading left to right like we do, the first information carries more weight than later information.
This helps explain why the first bits of information on a significant news story become so important. “A lie gets halfway around the world before the truth can get its pants on.”
Another error mechanism comes into play by means of insufficient adjustment from the reference number. We tend to adjust only until we reach the boundary of our uncertainty, and research has shown that such adjustments are typically insufficient. Take a sheet of paper and a pencil, and starting at the bottom, try to draw a line upward that is 2.5” long, without measuring. Then take another sheet of paper, and starting at the top, draw a line downward until only 2.5” remains. Typically, the line drawn upward from the bottom will be shorter than the space remaining below the line drawn down from the top. This is because we don’t know exactly how long 2.5” is, and we tend to stop when we reach the edge of our confidence range, not in the middle of it.
There are biases traceable back to anchoring that are related to the probabilities of conjunctive and disjunctive events. Conjunctive events are those where you are estimating the probability that multiple things, each with their own chance of occurrence, will all happen; “for my team to win the conference, they have to beat teams X, Y, and Z.” Regardless of how high the chances are of winning each of those contests, we tend to overestimate the probability that all will occur. Disjunctive events are those situations where you only need one of several things to occur; an example would be when you are trying to estimate the probability of a system failing when there are a handful of components, the failure of any one of which would cause the system to fail. We tend to underestimate the probability of disjunctive events.
There are still more error mechanisms – more than we’ll cover here. But hopefully you get the point: there are opportunities for our bias in more places than we realize.
How old was Gandhi when he died? A) Was he older or younger than 144? B) Was he older or younger than 9? Both options are, of course, ridiculous; we immediately realize that they provide zero helpful information. Are we going to use them as starting points for adjustment? Probably not. And yet, “primed” with this irrelevant data, a group of people asked Question A will still arrive at a significantly higher average age for Gandhi at death than those asked Question B.
As we saw in the example regarding judges being influenced by loaded dice, anchoring bias does not confine itself to laymen. Seasoned experts are affected as well. Research involving highly experienced real estate agents showed that they were greatly influenced by a home’s asking price, even when they were confident that they were not. Research also showed that business school students with no real estate experience were only marginally more affected by the asking price than the pros. The real difference between the two groups was that the students conceded that they were influenced by the anchor, while the professionals denied that influence.
It’s hard to fight something if you are unaware that it exists. So, congratulations! You’ve already taken the first step.
Right behind awareness is a second imperative that requires a little humbleness: acknowledgment that you are susceptible to it. This is perhaps the most difficult step; we think that human foibles are for other people, not us.
Step three is active fighting. This means shifting your thoughts out of autopilot into a deliberative mode. Assume that any number that is on the table has had an anchoring effect on you. If you are adjusting from an initial number, consider pushing your adjustment a little further than your comfort zone. One way to do this is to first adjust from the anchor to where you feel like you should stop; then deliberately go much further to the point where you want to dial it back; and then average the two endpoints. If your reference point is an extreme number from which you are adjusting, do your analysis a second time starting from an extreme reference number on the opposite side, and then average your results.
Also consider ways to actively unweight the earliest information. For example, consider reviewing all available information in the reverse order in which it was received.
Be mindful of our bias tendencies when evaluating conjunctive and disjunctive probabilities.
Finally, take time when you can. Time generally allows the gathering of more information, which can help to dilute the anchor.
There is a lot more to anchoring bias than is covered here, particularly in regard to marketing and salesmanship. There is abundant and eye-opening information available online on these aspects. But the focus here is not to help you get a raise or negotiate a better deal on a used car; it’s to help you be a better engineer.
Article Authored by David Aurand, QMS Manager