The Science of Bias
Human beings tend to incorporate unconscious and cognitive bias in just about every step of their decision-making processes. These biases tend to lead to systematic deviations from rationality, and often result in drastic oversimplifications in the way that we make decisions; which are known broadly as heuristics.
Here, we will discuss the scientific basis for these cognitive quirks.
What exactly is a bias?
In a single day, the average human being makes around 35,000 choices which can range from the route we take to work, to who we choose to employ [1]. In doing so, we make—unconscious or conscious—decisions that tend to neglect the rules of probability and sometimes even defy logic outright. These lapses in our judgement are known collectively as cognitive biases. While they can be useful in situations where we have limited time or copious amounts of information, they often lead to largely sub-optimal judgements [2]. Some common examples of these include the confirmation bias and the affinity bias [3]. The confirmation bias occurs when we focus only on information that confirms our existing beliefs while disregarding information that appears contradictory to those beliefs. The affinity bias, on the other hand, leads us to connect with others that we share similar backgrounds and interests with over those we do not.
Unconscious or Conscious?
Broadly-speaking, bias can be categorized in two simple ways: as unconscious or conscious. Unconscious biases are the kind that tend to exist most often in the workplace and recruitment process. They are biases that we commit unintentionally and without being aware of the fact that we’re doing it. As a result, they creep into our decision-making process in a way that is hard to control and even harder to prevent. Both the affinity bias and the confirmation bias are examples of unconscious bias. Conscious biases are, on the other hand, much more overt. In fact, these are the kinds of biases that we use intentionally. They are explicit and typically rooted in prejudice and discriminatory practices. Racism and sexism are among the many examples of conscious bias. While the kind of cutting-edge augmented intelligence technology we employ MeVitae has been designed to swiftly chip away at unconscious bias, we all still need to do our part to eliminate conscious bias from society.
Implicit Bias
In the workplace, one of the most common forms of bias that arises is referred to broadly as implicit bias. Implicit bias can be understood as a kind of unconscious social categorisation that results in our frequent sorting of people into groups that they display, often superficial, similarities with. Since the brain processes enormous amounts of information in rapidity, it employs regions like the amygdala (which handles emotive processing) to efficiently synthesize and categorise this information before we become consciously aware of it. In fact, when we come into contact with other people our brains instantly get to work sorting between people that are similar to us from those that are different [4]. This type of social sorting can be detrimental when it comes to recruiting new staff, especially with candidates of the opposite gender or of different ethnicities. It is also tends to go on behind the scenes and is scarcely something we are consciously aware of, much less can control on our own.
A brief history of bias
Implicit bias training began to take root in the years following the development of civil and gender rights legislation in the United States, and anti-discrimination laws across the world. In 1964, the US passed the Title VII of the Civil Rights Act, which prohibited employment discrimination on the basis of race, religion, gender, sex and nationality. But even while explicit prejudices were removed from the workplace, an implicit and historically-cemented bias remained. By 1998, an Implicit Association Test (IAT) was launched that aimed to determine which implicit biases were arising within organisations and during their hiring practices. While later findings brought much skepticism to the test, an industry emerged to sell “implicit bias training products” to corporations [5]. The problem is, we tend to not be aware of our biases when we express them, which is why we call them implicit. In fact, the very foundations of IAT were rocky – implicit bias training did not necessarily arise out of a desire for corporations to be more equal, but rather to help unequal corporations avoid litigation6. Implicit biases remain systematically engrained into society at almost every level, which means that dealing with them requires something that is not influenced by the same societal and behavioural norms as human beings.
The psychology behind bias
Conventional approaches that account for the origin of bias and heuristics include the cognitive-psychological perspective and the ecological perspective, both of which describe bias as arising largely out of the brain’s limited information processing capacity. While these perspectives have provided useful insights into why we deviate from rationality, they also employ circular arguments which rely on what is to be explained in the explanatory construct itself. In other words, they both argue that bias and heuristics result from limited processing capacities in the brain while contending that this limited processing capacity can evidenced from the fact that we use biases and heuristics when we reason. These prevailing viewpoints also fail to explain why heuristics and biases are typically consistent between different people and in different situations. This limitation is especially glaring when one considers that people vary enormously in their ability to perform different tasks and in the paths that they take to tackling these tasks [6].
The neuroscience of bias
More modern approaches to these problems attempt to rather account for the neural basis of how the brain defaults to heuristic-based decision making. From this point of view, decision making is a byproduct of the characteristics of information processing in the brain. The original design of these characteristics was to achieve perceptual motor control, so that we can physically react to that which we perceive, and to maintain biological integrity with our environment for the purposes of survival. In this way, biased decision making is the result of a gap between the original design characteristics of information processing and its application to more conceptual or analytical problems – such as those that we deal with in our everyday lives and especially in our work environments. One such example is, lateral inhibition. This is a process whereby an active neuron suppresses its neighbouring neurons’ ability to activate, which is especially useful for perceptual-motor functioning and for the blurry data that perception and motor controls use. The type of analytical tasks that we deal with every day on the other hand, such as selecting the best candidate for a given job role, requires an exact weighting of data and therefore a precise application of probability and logic. Lateral inhibition would be particularly poorly suited for this task and the level of accuracy that it requires. What’s more, these two tasks are processed in the same area of the brain (the cortical regions) which only further highlights the mismatch between the original design principles of the brain’s architecture and its application to our everyday tasks [7].
The associative brain
. Underlying all of the information processing that goes on in the brain, and therefore all of the decisions and actions that we make in the real world, is the concept of the association principle. This states that the brain—upon receiving any kind of information from the outside world—searches for patterns, links, and relationships with the knowledge that it has already. In fact, association is a basic principle that underlies just about any biological neural networks – i.e., animals and insects too! This simple property sits at the very root of the gender inequalities and racial disparities that exist in our society. It forms the core of our tendency to a) associate unrelated information, such as between a person’s physical characteristics and their ability to perform a job well; b) prioritise information that is compatible with our existing beliefs; c) retain information better left ignored, like the fact that we disliked the colour of a person’s shirt who applied for a role in our business; and d) to focus on information that is dominant while ignoring relevant information that wasn’t easy to recognise1. As a result, we tend to associatively classify people, events and ideas into categories that emerge out of stereotypical mixes of their traits and features.
The brain differs drastically from a computer in this respect, which stores information in a way that is indifferent to the characteristics of that information. The brain on the other hand selects, processes and integrates information not only through the qualities of the information it is receiving but also in terms of their compatibility with the brain’s current state and the connections between its neurons – which define that state. As a consequence, we see what we were expecting to see; and we are compelled to think, act and judge in ways that conform to our prior beliefs and knowledge. Biases are, in some sense, hard-wired into the brain itself. This makes the objective, data-centric approaches that we take at MeVitae all the more needed. While bias may be hard-wired into the brain, equality is hard-wired into our AI-driven solutions. These solutions provide real-time insights into your diversity data and company culture, while employing state of the art deep learning models and a Global Sourcing Engine that finds the right employees to fit that culture. We’ve also paired this technology with the world’s first anonymization API, which directly redacts bias from the resumes and cover letters of all of your applicants. In short, to best deal with the limitations of the human intelligence requires machine intelligence. Our philosophy is to put these two together so that you can capitalize on the best capabilities of both.
If you are interested in this service please contact us for more information.
Author: Riham Satti (Co-Founder and CEO)