Tottenham Report: Noise and bias lead to poor decision making By Andrew Tottenham, Managing Director, Tottenham & Co March 8, 2022 at 10:00 pm Tom Waterhouse, Chief Investment Officer WaterhouseVC.com, and I will be speaking at the International Casino Conference, part of ICE2022 on 11 April; our subject is, “Driving Better Investment Decisions for the Casino Floor”. As a prelude to the event, I thought I would give you some of my thoughts around decision making, not necessarily investment decisions, and how to get better at making them. Every day, we make decisions. Some can impact our health. How far away is that car? How fast is it going? If I cross now, will I get run over? Others are more mundane. Should I have ice cream with that slice of pie? Some are almost instinctive, while others require some thought, but that doesn’t stop some from using their gut reaction to make up their mind. The first thing to know about decision making is that we are human and fallible. Faced with the same decision, but with different and unrelated circumstances, we might make a different choice. Or given the same information, different people will make different choices. If an individual’s choices always tend in one direction, this is likely due to bias. If not, it’s likely due to noise. Secondly, whether we like it or not, we are all biased. Our inbuilt biases influence our daily existence and the decisions we make. Last time I counted, there were 18 different types of bias that infuse our decisions with prejudice. “Anchoring” refers to ascribing a higher value to something if we have just been exposed to higher numbers than if we were exposed to lower numbers. (Incidentally, this is why in a negotiation, it is better to name the price than let the other party go first; otherwise, you become “anchored” to the first number.) “Recency” is when we weight new information more heavily than old. Recruitment is an area where we fail miserably. Google “recruitment KPIs” and you will notice that post-hiring job performance rarely makes it to the list. The recruitment process in many companies allows for our biases to come to the fore. The name on the CV, the first impression when the person walks in, the handshake, what they are wearing, etc. all impact the judgment we make of a candidate. Unfortunately, it is rare that we can know what would have happened if we had hired somebody else for the position. It is unverifiable, so it is difficult to judge whether the recruitment process could be improved. Thirdly, we are easily affected by something that we have been doing. Back to recruitment: Did the interviewer have a dreadful commute that morning? Is he or she hungry (or hungover)? Studies have shown that judges give harsher sentences prior to lunch (when they are hungry, perhaps) and, unsurprisingly, doctors make poorer diagnoses and surgeons make more mistakes at the end of a long day. (Moral: Always see your doctor or have an operation first thing on a Monday morning.) How do we reduce bias? One approach has been to use AI as a tool. However, AI can and does have its own built-in biases. Plenty of research has shown shocking biases in AI involved in health-care and social-resource allocation. A cold, unemotional algorithm may seem to be better at making judgments, but the outcomes do need human oversight. Some research indicates that more intelligent people are worse at interpreting and implementing the outputs of AI. There is a sense of “that must be wrong” and they know better than the machine. Knowing that you are biased and what those biases are is not enough to overcome them. We believe that the world as we see it is the world as it really is and that most other people see the world the same way. Most walk through life assuming that their reality is the only reality and do not put any meaningful effort into questioning whether there may be other plausible realities. Careful — that way conspiracy theories lie (but I did say plausible). And we can be quite shocked when the person we have had a good working relationship with for a number of years expresses an opinion that shows they hold a completely different view of that reality. A long string of “successful” decisions allows people to become overconfident and to believe in their own infallibility. This is why it is rare for effective CEOs to stay effective in the longer term. They start to approach problems in the same way they always have, because it worked well in the past and should work well now. They fail to realise that time has moved on and things do not react in the same way. And experts (a group of people that I modestly believe includes myself) have more biases than the average Jo(e). How do we overcome noise and bias? I believe more care should be taken in devising a process that produces successful predictions for a series of similar challenges. Involve a diverse group to develop the decision-making process and to implement the process, but be careful of group think. Groups can get overly enthusiastic and reinforce one another’s belief that they are thinking well and have made or will arrive at the right decisions. Nobody likes to stand out by being the only naysayer in the group, so opposing opinions tend to keep quiet. We need to know what we need to know to support the decision and where and how to collect the data. Am I the only one who thought Donald Rumsfeld’s musings on the knowns and unknowns were actually among the more sensible things he said? Break the decision down into smaller chunks. What data do we need? How and where can we obtain it? What is the best method of analysing it? What does the outcome mean? Each person in the group should be able to analyse the data on their own and come to their own conclusions before discussing it with the group. Group discussions about analyses before each group member has reached their own conclusion will rapidly lead to members coalescing around the most dominant member’s conclusion. Once the method has been tested and the outcome against prediction known, the group should revisit it and see if can be improved, even if the prediction was right. When judging how good the procedure actually was, many will look at the prediction and, if the outcome matches, give themselves a big slap on the back, after all, their process produced an answer that was borne out in practise. Instead, they should be seeking a decision process that produces the best outcomes in a number of similar cases. When asked to pick an ace of spades from a deck of cards, an ape can sometimes get it right! Hopefully, the above can set you thinking about what a good decision is and how it can be achieved.