Noise and decision making

Helping you be more objective

Nikki Matthews

Senior Investment Consultant

People are not the same at all times and behave differently in certain situations, in different roles and this may even change over time. Further details can be found in my previous article How do you decide? - LCP Vista Spring 2021.

It is important to remember this point when judging people or fund managers based on very limited interactions revealing only certain aspects of their behaviour, which may be heavily influenced by external factors. My article covers the key points from the book 'Noise'.

Large discrepancies have been found in sentences that judges give for identical crimes.

The resulting differences are predominantly due to the identity of the person doing the judging, their personal values, personality and life experiences – some judges are harsher, others more lenient. In some cases this could be partially related to bias against the accused, but largely the difference in sentencing has nothing to do with the case being judged! This is variability between judges.

There is also variability within judges ie the same judge could pass a different sentence for the same crime on a different day – this has been found to be influenced by external factors, such as the weather (studies have found that on cloudier days admissions officers paid more attention to the academic achievements of candidates and on sunnier days extra-curricular activities were more heavily weighted) and internal factors (the judge’s mood and how hungry they were feeling).

The unwanted variability in judgements is noise. This is different to bias which exists when most errors of judgement are in the same direction. Even removing all bias, noise will remain due to random variability and inconsistencies in judgment which lead people to evaluate the same information differently. On average noise should cancel out – the issue is that it does not cancel out for the same issue/case and therefore outcomes are unfair.

Guidelines can help reduce noise but in most cases they allow a higher degree of judgement than is optimal.

Large discrepancies have been found in sentences that judges give for identical crimes.

What next?

So how can you be more objective?

Having multiple meetings before making important decisions, for instance meeting the people managing a Fund several times and under different settings can help form a more holistic view of the people managing it.

Objective ignorance

The process of trusting your gut based on an internal signal is often wrong, though it has great emotional reward when it is correct (which does happen sometimes), and is therefore difficult to relinquish. The judgement is often wrong as it is based on limited information.

In theory adding extra opinions and aggregating judgments should reduce noise, however group dynamics can further exacerbate noise, due to influence within the group – if an influential sceptic voices his opinion first this may lead the rest of the group. Initial popularity can greatly influence the decision by shaping others opinions.

Independence

Independence within groups is key to reducing noise or alternatively, extremely favourable circumstances in which people share their knowledge and everyone is equally receptive to another’s opinion. This requires having no preconceived notions about people in the group, which becomes harder the more familiar the members of the group are to each other.

A way to decrease this is to share opinions anonymously and decide the order of discussion at random. Obtaining information from the initial source can be beneficial as often the ‘second’ person passing information on has been influenced by the first person’s view and the information is not necessarily objective. Breaking down the decision into smaller tasks and scoring each one on several factors can also be beneficial as it makes decisions more objective.

How to improve?

Having a checklist with sources of noise and bias may help in making better decisions as you would be more astute to making errors of judgement.

AI may help in standardising decisions to a greater extent, indeed machine learning algorithms can even be used to predict decisions of the US Supreme Court based on data. AI can also spot patterns in the data that humans may not be able to detect, however it is important to remember that the output is only as good as the input and rules specified by humans, but it can be hugely beneficial to have an additional objective input to aid decision making.

Case study

Fair value of a security is difficult to estimate, differences amounting to 41% were found between analysts working in the same company that were provided the same information. Fair value is subjective, noise overall ‘cancels out’ where markets are efficient, however there are winners (those who pay a price below fair value) and losers (those who pay a price above fair value). Where markets are not efficient, the fair value is wrong. So what can you do? Use several valuation methodologies to determine the value of an asset – this will give you a range of possible outcomes; averaging may help in getting closer to fair value and at least you will not be surprised if the valuation turns out to be different to the price you paid.

Conclusion

It's important to acknowledge the inherent variability in human behaviour, as individuals act differently under different circumstances. Making comprehensive assessments about someone's character or competence from limited observations is likely to be flawed, your opinion may also be impacted by your own mood at the time; where possible it is beneficial to develop opinions and make decisions over time.