Photo: www.jamesrobertwhiteart.com
Background
In biology, every school child soon learns (and then forgets?) the idea of ‘survival of the fittest’. This phrase is due to Darwin’s theory of natural selection but actually originated in the writing of Herbert Spencer. What people tend to think this means, once they have forgotten their biology lessons, is that the organisms that are the ‘strongest’ are the ones that will survive. Unfortunately, this grossly simplistic version was also used to support the idea that ‘might is right’ and contributed to the terrible idea of applying inaccurate versions of Darwinism to society. One example, was supporting imperialist ideas in the late 19th and early to mid-20th century based on the idea that some ‘races’ were superior to others genetically. Another terrible case lead to eugenics, the idea of selective breeding of humans. This shamefully involved forcibly sterilising some people that the state decided should not be allowed to reproduce.
However, putting all this terrible history aside, and returning to the actual science rather than corrupted versions of it, the way the phrase was originally used by Spencer, and then Darwin and other evolutionary biologists, is that ‘fittest’ meant simply ‘most suited to the environment’. This was meant as a way to explain how evolution worked. It was meant to explain that the organisms that would enjoy greater reproductive success were those best suited to whatever environmental conditions prevailed at the time.
False Fitting in Biology
However, in recent decades, modern biology research into evolution has shown examples where an organism or a community of organisms enjoying greater reproductive success in the short-term leads to lack of fitness to the environment in the medium or longer term. As a result, we see examples like this, of 'false fitting'. By understanding these cases of false fitting, businesses can better avoid solutions that also provide 'false fits' to their problems.
In business and government, commentators have bemoaned ‘short-termism’ for decades. This is the idea that organisations and leaders look to short-term solutions and ‘quick wins’ at the expense of the medium and longer term. Using biological research to help us to understand the problem potentially allows for more clarity and detail than has been available to other commentators, up to now. One example of this, is Turing Meta’s work on the suprisingly complex consequences of this type of analysis for economic policymakers.
Why the claim for originality and new insight? Well, biological organisms face the same problems of how to promote growth, adaptation and deal with uncertain environmental conditions. We should therefore expect that biologists studying organisms are able to clarify such issues in a way that helps businesses and organisations to better understand their challenges, too. And indeed this is exactly what has happened. Translational research by Turing Meta is then able to 'tap into' significant levels of expertise and significant levels of control over the experimental subjects, by biologists. This level of control over biological experiments is simply unavailable to economists and social scientists who are then unable to provide the same level of experimental control and rigor. Turing Meta translates clever biological analysis of these longstanding issues to provide more clarity and depth of insight. Turing Meta utilises this modern biological research which was carried out for the public good. As this biological research was already funded by your tax money, why shouldn't your business or organisation directly benefit?
Biology case studies
One example of this is research into cooperation between bacteria. There are some bacteria that produce a chemical that allows them to eat iron found in the environment. However, experimental biologists engineered a mutant that doesn’t produce the chemical itself. This greedy bioengineered mutant can still eat the iron that its neighbours are doing work to allow them all to eat, however. In other words, these mutants are now ‘free riders’, meaning they benefit from work done by others without doing their fair share of the same work. Because they don’t need to use energy to produce the chemical like their neighbours, by 'survival of the fittest', these mutants will now divide and reproduce more quickly than their harder working neighbours, but this is a ‘false fit’.
Eventually the whole colony is full of mutants that can’t now produce the chemical they need to eat iron, and continue to live. As a result in the medium or longer term, the mutant was a was not really best fitted to the environment. Their temporary success crowded out the better fitting neighbours and eventually killed off the whole colony. In this case, 'survival of the fittest' actually lead to extinction.
Other examples of this exist in the research literature. Another case is where natural mutations produce a false fit to the real environment by being produced when bacteria are being grown in a lab's petri dish. Random mutations are likely to disable costly metabolic processes that are needed in the wild, again, to convert different chemicals to food. But these metabolic processes are irrelevant to the petri dish environment where the food is not available. Bacteria cells with these random mutations are now a better fit in the ‘short term’, because they need less energy to reproduce. Again, it is 'survival of the fittest'. But this is advantageous only in the petri dish environment, which they now fit to well. When released into the wild, however, they will quickly perish as the metabolic processes they now need again, to eat a greater variety of food are no longer available, because their natural mutations disabled them. Short-term 'survival of the fittest' was only achieved by ‘cannibalising’ the fitness level with future environmental states just as in the previous, example where the biologists had to engineer a mtuation for the same effect.
Applying False Fit Analysis to your business problems
Your business can benefit from this insight by analysing and checking your analysis for signs of false fitting. A good analyst will recognise the risk of overfitting or underfitting a model to data. However, another key skill for the modern analyst, especially when making recommendations, is to analyse whether the model is potentially producing a false fit, too, directly analogous to the way the petri dish was a simplified version of the real enviornment.
This means that when you apply a recommendation there is some risk that you appear to significantly increase, say the efficiency of some process, or the productivity of some team. However, if the increase would appear to be significant, there is a risk that this is a false fit, just like free riding mutants, or bacteria grown in an environment of the petri dish, which isn’t a good model of a real environment, outside of the model (your cognitive model or metrics that represent the problem are like the petri dish).
One example of this might be simplifying a business process to make it more efficient. It may be that the process is only now more efficient because it ignores rarer, edge cases that are actually very significant in their impact, such as the cost of missing certain steps as a risk of a costly error in a minority of cases. This is directly analogous to the petri dish example, where the model of the process is missing essential details. As a result, it is dangerous to choose the ‘best fit’ to some simple measure of efficiency, such as the cheapest supplier, because this increases the risk that you are actually selecting a false fit. Rather than a bias to eking out the last 3% of cost savings, businesses should be suspicious of the cheapest supplier or the most ambitious plans and targets.
Another example, might be the analysis of where savings can be made in testing or longer terms design of some technology project. It may be that such savings to the project budget are genuine, but if they appear very significant, it may represent a false fit, where you are only saving the highest amount of energy by also cannibalising returns in the medium or longer term. The way to avoid this depends on understanding the likely costs to medium and long term results and modelling this in the analysis of where savings can be made, and by correcting the bias for the cheapest supplier or plan. Turing Meta can help with this analysis, and create methodologies for selecting from options which reduce the odds of false fitting. Future posts will also go into detail on how this can be done, but essentially, we can think of false fitting as a stastically higher likelihood for options that appear to offer dramatic, or simply, the highest cost savings or efficiencies. We should therefore be systematically biased against the highest efficiency option unless we are very confident our model is complete in all details.
In the simplest case, to apply this principle:
Ask yourself the next time you are selecting from several different options whether the cheapest option, most efficient result, or highest quality candiate is also more likely to be a 'false fit'.
Analyse the differences between your model the the real thing and make the simplifications explicit. This helps to understand the risk of false fitting.
Also, share with management the concept of false fitting to help to explain the thinking behind not querying what appears to be the cheapest or most efficient option.
When hiring for instance, be aware that testing people is statistically more likely to generate a false fit, if you rely on the highest score to choose between candidates. This is because the test is a vastly simplified model of real work. Claims that tests correlate well with performance in actual work, cannot adequately consider the candidates who did not get selected, because social scientists cannot use the same level of experimental control to derive their results and support their claims, when compared to biologists.
Subscribe for more updates from Turing Meta to help with your future analysis and decision-making.