An Essay on implicit bias.
The aim of this essay is to establish the nature of a bias. In the first half of the essay, we will show it is possible to hold that people have implicit biases. We will do this by giving a definition of what an implicit bias may be, and by providing a case study to support our position, as well as by showing how an implicit bias may be different from an explicit bias.
Following this, we shall provide the argument that holds that a bias is a type of belief and so the bias itself can be implicit or explicit.
We shall then provide a response to this argument which would be the position we favour that holds that biases are not themselves the mental states (which can be implicit or explicit) but rather are traits which are formed from mental states such as beliefs.
SECTION 1: WHAT IS AN IMPLICIT BIAS
According to Micheal Brownstein, we can define implicit bias as the automatic features of prejudiced judgement and social behaviour that may be relatively unconscious (Brownstein, 2017). We see cases of implicit bias in scenarios where individuals make judgements about and act towards differences in race, and gender. There has been a development and increase of empirical support for claims that people despite their conscious beliefs have implicit biases (Brownstein and Saul, 2016). For instance, case studies have indicated that participants that take part in shooter bias tasks are more likely to shoot unarmed black men in a simulation than unarmed white men (Brownstein and Saul, 2016). Even more so researchers discovered that these same participants are more likely to fail to shoot armed white men than armed black men (Brownstein and Saul, 2016).
In particular, the Implicit Association Test shows through an evaluation of implicit attitudes that people may have implicit biases (Brownstein and Saul, 2016). In this study, the subjects are told to sort pictures and then words at various intervals into categories as quickly as they can while making as few errors as possible (Brown- stein and Saul, 2016). Participants are first presented with variations of four images, two of European Americans and two of African Americans, and told to sort the pictures to the appropriate identifying response either on the left or the right (Brownstein and Saul, 2016). Then they are told to sort words such as ‘pain’ or ‘happiness’ to the corresponding appropriate response on the left or right being either ‘good’ or ‘bad’ (Brownstein and Saul, 2016). Finally, participants are to do the same task again but this time words may follow after a picture and vice versa as opposed to having just pictures or words appear in one go (Brownstein and Saul, 2016). Additionally, in the final section, the corresponding responses could switch positions. That is, if originally the appropriate responses were arranged such that “European American” and “Bad” were both on the right while “African American” and “Good” where on the left, it would be re-arranged such that “Bad” and “Good” switch places while the others stay the same (Brownstein and Saul, 2016).
This research revealed that 70% of white participants would be faster and make fewer mistakes when ‘African American’ and ‘Bad’ are on the same side (Brownstein and Saul, 2016). Researchers state that this represents an implicit preference for white faces over black faces (Brownstein and Saul, 2016). Also, the findings revealed that 40% of black participants have an implicit preference for black faces, 20% of black participants showed no preference and 40% of black participants showed a preference for white faces (Brownstein and Saul, 2016).
This study shows us that we have implicit attitudes towards particular social groups regardless of whether we are a part of that group or not. The major point we take from it is that implicit attitudes possibly exist. Although, regardless of its results, this research fails to show us how we may distinguish an implicit bias from an explicit one. The study also does not give concrete evidence for why we may believe that it is the bias identified that is implicit. So we hold that it is necessary to have an ac- count that explains the nature of implicit bias and how it may exist separate from an explicit bias, and that is why in the next section, we shall consider such an account.
SECTION 2-IMPLICIT BIASES AS DISTINCT FROM EXPLICIT BIASES
The Dual Process Theories model offers explanations which allow us to understand how an implicit bias may be different from an explicit one. Dual process models, influenced by dual process theories of mind (Brownstein and Saul, 2016), define explicit and implicit attitudes in terms of operating principles (Brownstein and Saul, 2016).
The term ‘operating principles’ in our view simply means psychological processes of the mind such as learning and the formation of judgements. Both models propose that there are two operating processes, the associative and the propositional processes which depend on an agent’s assessment of the truth of a given interpretation (Brownstein and Saul, 2016). In other words what these models are proposing is that the mind comprises of two streams for processing information.
The first stream is the associative stream which according to these models under- goes ‘Type 1’ processes (Brownstein and Saul, 2016). This stream is automatic, non- conscious, fast, and associative (Brownstein and Saul, 2016). It is also made up of a slow learning memory system that is responsive to social conditioning and experience (Brownstein and Saul, 2016). Therefore, it would be fair to say this stream would be the stream through which we develop implicit beliefs and thus implicit biases.
Meanwhile, the second stream according to these models is the propositional stream which undergoes ‘Type 2’ processes and is slow, conscious, controlled and rule-governed (Brownstein and Saul, 2016). Furthermore, it has a memory system that is capable of ‘one shot’ learning in response to explicit tuition (Brownstein and Saul, 2016). Here we would hold that this stream is more responsible for the development of explicit attitudes
Following this proposition, on a mundane level, it could be the case that in a person’s mind, two main cognitive processes are going on which in turn determine the conceptions that guide a person’s behaviour towards another person or thing.
Think about the case of a person with a mental illness such as depression or anxiety. Following the suggestions of these models, we could argue that such patients are dealing with an inability to have ‘Type 2’ processes dominate their cognitive process- ing consistently. Such an individual probably finds themselves in automatic, non- conscious, associative reasoning that negatively affects their self-perception as well as their perception of the world around them. Also because the cognitive process that occurs in this reasoning is made up of a slow learning memory system that is responsive to social conditioning and experience we can see why it is so difficult for one to break whatever resulting conceptions that they easily develop. If there is a persistence of the environment that allowed the depression or anxiety develop, then we can see how separating oneself from the beliefs that formed in that environment would be hard.
Moreover, we know that this is a case of depression and not just a case of one who just doesn't like themselves or the world around them, because of moments of clarity. What we mean here is that there are moments where according to this model the person's 'Type 2' reasoning activates, and they become aware of the fact that their conceptions at the time where their 'Type 1' depressive reasoning was operating are not beliefs that they consciously endorse. The depressed agent does not want to be depressed or endorse the beliefs they develop in a depressive episode; even there are cases where a depressed person does not know they are depressed. This position, therefore, signals that depressed reasoning could be more implicit than explicit and that the dual process theories give us the means to explain the difference. In the next section, we shall provide more of the arguments that claim that we should view implicit biases as beliefs, and thus hold that it is the bias that is what may be implicit or explicit.
2.1-IMPLICIT BIASES AS BELIEFS
Consider arguments made by folk psychologists which state that there are two kinds of beliefs of which are implicit beliefs and explicit beliefs (Frankish, 2016). According to these psychologists, an implicit belief is said to guide behaviour and thought with- out being recalled consciously, while explicit beliefs are supposed to require conscious recalling (Frankish, 2016). Following this premise, we hold reflective behaviour as any behaviour that occurs as a result of conscious recalling, while unreflective behaviour is any action made without any conscious recalling.
For instance, some people in the United States instinctively react as though they feel safe around police officers, while others, particularly minorities, feel threatened in the presence of an officer. Arguably these reactions are a result of the beliefs these two types of people have about officers. A middle-class white man is more likely to believe that the law is on their side than a black working class boy. We believe this is a fair assumption to make considering first of all the evidence that there is an existence of biases towards black people as shown in the case study in section 1 and just generally from how popular culture and the news has depicted race relations in the United States over the years. Returning to the point, according to the folk psychologist, if the black boy does not consciously endorse the belief that officers pose some danger then that belief is an implicit belief but if it is consciously endorsed then that belief is an explicit belief. Further, according to the dual process theories, we would know if it is an implicit belief if the type of cognitive process that occurs is a 'Type 1' process and we would know if it is an explicit belief if the cognitive process that occurs is a 'Type 2' process.
This case also applies to the scenario of the depressed agent. Equally, as we have established, it seems to be the case that the conception the agent develops regard- less of the cognitive process that occurs would be a belief. Whether or not it is an implicit or explicit belief would depend on the type of cognitive process that leads to this belief.
In sum, the dual process theories model explains how implicit biases are developed, and thus how we can evaluate if a bias is implicit or explicit, and the folk psychology model tells us what an implicit bias is, which is a type of implicit belief. Together they give an account of the nature of implicit bias that although is informative, is quite in- sufficient. In the next section, we shall refer to the combination of these models as the 'unconsciousness account of implicit bias', and we will show why this account is an insufficient account of implicit bias.
SECTION 3-THE PROBLEM WITH THE UNCONSCIOUSNESS ACCOUNT OF IM- PLICIT BIAS
While we hold that there is a measure of unconscious cognitive processing which is responsible for the formation of implicit beliefs and thus bias, we do not accept that the lack of conscious awareness or endorsement of a bias is sufficient to distinguish an implicit bias from an explicit one.
Consider this scenario where, Juliet, a philosophy lecturer, knows that there is no scientific evidence for racial differences in intelligence and sincerely argues for the concept of equality of intelligence, but in her unreflective behaviour and judgements we find that she has a systematic racial bias (Frankish, 2016).
Juliet is more likely to view a white student as bright while not necessarily thinking the same of black students. She is also more inclined to be surprised when a black student makes an intelligent comment than when a white student does the same. Juliet behaves and judges as though there are racial differences in intelligence, though she does not think that there are racial differences in intelligence and if that thought occurs, she rejects it (Frankish, 2016). According to the unconsciousness model, we are to hold this case as a case of Juliet having an implicit belief and so an implicit bias about racial differences in intelligence.
Against this proposition would be the objection that it is not the case that Juliet has an implicit belief but rather what we see is Juliet attempting to deny her explicit beliefs. Moreover implicit biases do not exist but rather the assumed case of implicit bias is a state of an agent self-deceiving about their explicit bias (Mele, 1983). To self deceive is an act of acquiring a false belief because of the desire to have that belief (Mele, 1983). The false belief, in this case, would be Juliet's belief that she is not biased when it comes to determining if there are racial differences in intelligence. Ac- cording to the self-deception perspective a person cannot belief p and -p (not p), so if that person asserts p and then acts in a way that is -p then that person believes -p and has acquired a false belief about p (Mele, 1983). So Juliet cannot explicitly belief that she is not biased and still have implicitly biased beliefs but rather what is the case is that Juliet has explicitly biased beliefs which she has acquired the false belief about that they do not exist. Therefore what she actually believes is that there are racial differences in intelligence and her assertion that she does not have this belief it a case of Juliet deceiving herself.
To further cement this argument is the position that the unconsciousness account fails to give us a reason why an agent who is conscious of their behaviour being affected by Type 1 processes would be unable to control this effect by giving their Type 2 cognitive processes dominance in operating how the agent learns and forms judgements. Furthermore, it is hard to believe that one who accepts egalitarian ideals would not be able to tell when their behaviour is departing from these ideals.
SECTION 4: HUEBNER'S SOLUTION
The argument in the previous section shows that the claim that states that we may be able to distinguish between an implicit bias and an explicit bias is faced with a compelling objection that is the self deception objection. If this objection succeeds then it follows that an agent cannot have an implicit bias that is distinct from an explicit one. Consequentially, this means that an implicit bias is not a belief as beliefs can be implicit or explicit.
In his account, Huebner potentially gives us a solution to this problem by identifying how the aggregate voting patterns of three cognitive systems in an agents mind can lead to the formation of either an implicit or explicit bias.
Huebner holds that there are three systems within which learning, judgement and memory formation takes place. These three systems are; an associative Pavlovian system which triggers a withdrawal and approach reaction, an associative model- free system which reflexively assigns values to actions based on the result of previous actions and a reflective model-based system which generates decisions that are forward-looking (Brownstein, 2017).
Huebner’s argument is that there is a parallel functioning of these systems and that many of our decision-making and resulting actions depend on contributions from these three distinct types of systems (Huebner, 2016).
First, he argues that the Pavlovian systems in our cognitive processes are responsible for tracking things which an agent might consider as threatening or dangerous (Huebner, 2016). These systems get information by examining the association an object has with being threatening (Huebner, 2016). The system then creates from the information given, predictive algorithms by calculating these potential threats (Huebner, 2016). Furthermore, if a subject is concluded to be dangerous, then the Pavlovian system would guide an agent to avoid and be averse to it (Huebner, 2016). Social interactions can, therefore, guide the evaluation of the levels of potential danger that comes from a subject, this is because all a Pavlovian system needs is an association (Huebner, 2016). Moreover, the more associations that imply that a subject is potentially harmful, the higher the resulting guidance to avoid and be averse to that subject. Therefore if an agent is surrounded by internal structures such family or friendship groups and external structures such as the media which associates groups of people with potential danger, then it follows that that agent will be inclined to avoid these groups.
To support the claims about the functioning of Pavlovian systems, Huebner states an example of case studies that show that there is increased activity in the amygdala when researchers expose white participants to the face of a black person (Huebner, 2016). This revelation is relevant because recent empirical data suggests that the amygdala is a centre for the evaluation of both social and biological prominence of stimuli in manners that track abstract patterns of risk and reward (Huebner, 2016). More importantly, researchers found that there is a generation of an avoidance signal that occurs in the amygdala based on risks associated with biologically and socially significant stimuli (Huebner, 2016). Consequently, for Huebner, the reason behind the increased activity in the amygdala when researchers exposed white participants to black faces is because participants are reflexively evaluating potential dangers, risks and indicators of trustworthiness (Huebner, 2016).
Next Huebner states the role of Model-free systems. He argues that these systems have the job of checking the experiences of an agent for errors in past predictions (Huebner, 2016). There is a resulting adjustment of erred predictions when experiences of rewards and punishments suggest that they are flawed (Huebner, 2016). However, these systems are easily affected by what Huebner refers to as fictive-error signals which are signals that compare actual outcomes of actions or events with what an action or event may have been (Huebner, 2016). Furthermore, while these signals could provide a way through which model free systems can test action before they happen, they unfortunately also compete with the processes that the model-free system already undergoes to check for predictive errors (Huebner, 2016). Further- more, if an agent is not careful, these fictive error signals could contribute to the weakening or strengthening of perceived associations, which in the former case would allow for an overpowering of Pavlovian evaluations or even a flawed conclusion about what may or may not be an erred past prediction (Huebner, 2016).
To support his claims about the model-free system, Huebner makes reference to tests which have shown that when an agent seems to be, in her cognitive processes, conducting a model-free task, we find activity in regions of the brain such as the me- dial prefrontal cortex (Huebner, 2016). This finding is relevant because that is the region of the brain that we associate with abstract social reasoning (Huebner, 2016).
Moreover, an implicit bias would, therefore, be a result of an overpowering of the model-free system by the Pavlovian system, which we have identified to be an easy source of the formation of implicit bias. Alternatively, it could be the case of when the model-free system attempt to check Pavlovian results it would fail to find errors in reasoning because fictive-error signals alter its results (Huebner, 2016).
Finally, Huebner states that we are not just machines that respond to stimuli as the first two identified systems may suggest (Huebner, 2016). Furthermore, evidence suggests that agents often construct counterfactual models of the world around them, which they then evaluate, allowing them to predict the consequences of actions before they act (Huebner, 2016). This case, he argues is a result of the existence of the model based system (Huebner, 2016). The reasoning that occurs as a consequence of this system is one that allows agents to build decision trees which they use to represent the value of action outcome pairs (Huebner, 2016). Agents search these trees to discover which actions would, therefore, be likely to present the best overall result (Huebner, 2016). This process, Huebner argues, relies on the mechanisms in the prefrontal cortex that represent values and stored goals in work- ing memory (Huebner, 2016). Moreover, while the decisions that come about as a result of model-based reasoning are more accurate and more flexible than the ones produced by the predictive Pavlovian and model-free learning systems, the operations of model-based systems are more costly with regard to neural time and re- sources (Huebner, 2016). This causes a conflict in the decision to use the costly model based system or the simpler associative systems (Huebner, 2016).
Altogether, this account not only shows how implicit biases may be formed but also give a better account of how they may exist separate from an explicit bias. Implicit biases are a result of the processing of flawed associations in the Pavlovian systems. Alternatively, they are a result of the poor influence of a Model-free system but fictive error signals. Meanwhile, an explicit bias would be the result of the functioning of all mentioned systems, particularly, the model-based system as it is the one system that engages with more tangible external information the most.
We can see how the first two mentioned systems could dominate the cognitive processes of the mind as they as less costly to use and are probably quicker in enabling the learning processes or formation of judgements in an agent. We take implicit bias according to this model as then being just a result of flawed reasoning due to the collection of incorrect associations. This account also bypasses the self deception objection simply because it shows that there is a difference between implicit biases and, ones and thus that it is possible to belief two things at once.
However, we propose that what this model shows does not necessarily defeat the self deception objection because it does not show us if one can have an implicit be- lief and an explicit belief at the same time, rather what is shows us is how we may form biases.
SECTION 5-THE TRAIT PICTURE ACCOUNT OF ATTITUDES AS A SECOND OBJECTION
The Trait Picture account supports the conclusion in the last section as it argues that attitudes are traits and unlike mental states, traits cannot be occurrent (Machery, 2016). This means that traits, such as bias or courage, are not things that occur and have specific effects when they do occur (Machery, 2016). Rather traits depend on mental states, which are occurrent, in order to exist (Machery, 2016).
Also the Trait Picture model argues that indirect measures of attitudes such as the IAT do not measure attitudes themselves but the psychological bases of an attitude and particularly components that are not accessible to introspection (Machery, 2016). Since, attitudes such as biases depend on non-propositional associations be- tween concepts, then tests such as the IAT are measuring the strength of these associations rather than the biases themselves (Machery, 2016). So if attitudes are traits, i.e., dispositions that exist through a dependence on mental states, then it follows that attitudes such as biases are not what can be implicit or explicit but rather it is the mental state that creates biases that can be explicit or implicit (Machery, 2016).
Attitudes are traits because they are broad track dispositions to behave and cognise in a way that indicates a kind of preference (Machery, 2016). Furthermore, to have a bias towards a person or thing, is to be disposed to relate that person or thing with negative evaluations of what they are (Machery, 2016). Also, as we have stated traits have a psychological basis, therefore an attitude such as a bias towards black people would depend on a moral belief, and that is the moral belief that racism is right or on emotions such as a fear of black people (Machery, 2016).
This model finds support for the claim that its propositions about what an attitude may be are a more accurate account of attitudes in the fact that it holds that empirical testing shows a weakness in other accounts that does not affect its account. Ac- cording to this account, there are four reasons why models that attempt to explain biases as being mental states which can be of an implicit or explicit form are not the best accounts of what an attitude may be (Machery, 2016). These reasons are in the form of findings which identify inconsistencies between the claims of accounts that holds an attitude as being a mental state and the correlation between results of empirical testings. The four findings determined that in empirical testing there are, low correlations between indirect measures, variations in the correlations between indirect measures, variations in the indirect measures of attitudes, and a low predictive validity of indirect measures (Machery, 2016). Due to limited space we shall only focus on the first finding and evaluate whether or not this finding may be able to show that the Trait Picture model may be a better account of the nature of attitudes.
Sherman et al. reported correlations between several forms for implicit association research studies as well as studies of evaluative priming varied from -0.11 to 0.11.
These correlations are thus insignificant and low which is puzzling if an account of implicit attitudes such as Huebner's is to be correct (Machery, 2016). This case is puzzling because we are to expect larger correlations than what has been found if the indirect measures of attitudes discover the same kind of implicit attitudes in participants (Machery, 2016).
In contrast, the Trait Picture model arguably predicts this result of low correlations (Machery, 2016). According to the model, indirect measures tap into different components of the psychological bases of attitudes (Machery, 2016). Therefore there is no reason why we are to expect that these components would correlate with each other (Machery, 2016).
CONCLUSION: THE NATURE OF BIASES
We started this essay by providing the arguments that hold that we may have implicit biases because implicit biases are a type of implicit belief.
We responded to this claim that it is not the case that an implicit bias is a type of be- lief because biases cannot be implicit and explicit at the same time, rather what is assumed to be an implicit bias is actually a case of self deception about an explicit bias.
Next we provided a possible response to this claim which was that Huebner’s ac- count shows us that implicit biases can be developed and be distinct from explicit biases because of the nature of the cognitive systems in people.
We responded to this claim by stating that while Huebner’s account does tell us something about the nature of biases what it shows us is how biases may be formed and not that biases can be implicit or explicit.
Finally we argued that we get the best picture of what a bias may be from the Trait Picture model.
M. (2017). Implicit Bias. [online] Plato.stanford.edu. Available at: http://plato.stan- ford.edu/archives/spr2015/entries/implicit-bias/ [Accessed 8 May 2017].
Brownstein, M. and Saul, J. (2016). Implicit bias and philosophy. 1st ed. Oxford: Oxford University Press.
Frankish, K. (2016). Playing Double. In: M. Brownstein and J. Saul, ed., Implicit Bias and Philosophy, Volume 1: Metaphysics and Epistemology, 1st ed. [online] Oxford Scholarship Online. Available at: http://www.oxfordscholarship.com/view/ 10.1093/acprof:oso/9780198713241.001.0001/acprof-9780198713241-chapter-2 [Accessed 8 May 2017].
Huebner, B. (2016). Implicit Bias, Reinforcement Learning, and Scaffolded Moral Cognition. In: M. Brownstein and J. Saul, ed., Implicit Bias and Philosophy, Volume 1: Metaphysics and Epistemology, 1st ed. [online] Oxford Scholarship Online. Available at: http://www.oxfordscholarship.com/view/10.1093/acprof:oso/ 9780198713241.001.0001/acprof-9780198713241-chapter-3 [Accessed 9 May 2017].
Machery, E. (2016). De-Freuding Implicit Attitudes. In: M. Brownstein and J. Saul, ed., Implicit Bias and Philosophy, Volume 1: Metaphysics and Epistemology, 1st ed. [online] Oxford Scholarship Online. Available at: http://www.oxfordscholarship.- com/view/10.1093/acprof:oso/9780198713241.001.0001/acprof-9780198713241- chapter-5 [Accessed 14 May 2017].
Mele, A. (1983). Self-Deception. The Philosophical Quarterly, [online] 33(133), pp. 365-377. Available at: http://www.jstor.org/stable/2219163?seq=1#page_s- can_tab_contents.
Payne, B. and Gawronski, B. (2010). Handbook of implicit social cognition. 1st ed. New York, N.Y: The Guilford Press, pp.366-367.
There are many words in English that despite having very similar sounds have completely different meanings. This can lead to confusion and usage problems for native and non-native speakers alike, and the words implicit vs. explicit are no exception to this.
These two words have almost opposite meanings but are regularly confused because of their similar sound. Since they do have such different meanings, you want to be sure you’re using the right one. In this post, I want to go over the definitions of these words, explain their differences, and have you take a quiz on their meanings.
After reading this post, you won’t ever again ask yourself the question, “Is it explicit or implicit?”
When to Use Implicit
The definition of implicit is, “implied or understood though not plainly or directly expressed.” Something is, therefore, implicit when it is not directly stated but is either suggested in the wording or necessary to effectuate the purpose. For example,
- There is a morality implicit in his writings.
- She implicitly said she likes white shoes by saying she likes all colors but tan.
In the first example, the writer may not have clearly or directly laid out a moral vision, but it is understood through the characters, their actions, and their experiences.
In the second example, the woman states that she likes all shoe colors but tan. While she doesn’t directly say she likes white shoes, she implicitly does because white is not tan.
When to Use Explicit
The definition of explicit is, “to fully and clearly express something, leaving nothing implied.” Something is explicit when it is cleared stated and spelled out and there is no room for confusion, as in the writing of a contract or statute. For example,
- The law was explicit in whose tax rates were to be raised.
- He said explicitly, you will not attend that concert.
In both of these examples, the word explicit is used to demonstrate something that has been clearly and unambiguously expressed or stated. There is no room for doubt because everything is clearly and directly communicated.
This is what separates these two words. Something is implicit when it is implied but not directly stated. Something is explicit when it is directly stated and leaves no room for uncertainty.
Quiz and Sentence Examples
- The speaker’s intentions were not made ______.
- The students found an ______ political statement in their teacher’s remarks.
- Let me be ______, I do not support this.
- We have not finalized the decision, but have an ______ agreement.
Display the answers below.
Tricks to Remember
Here is a handy trick to remember the difference between these words. Remember this and you won’t ever fall short when thinking, “Is it implicit or explicit?”
A good way to keep explicit vs implicit apart is to remember that Implicit is an Implied or Indirect statement. Both of these start with the letter “I.”
Explicit starts with an “E” and is Spelled Out, so there is no confusion.
Implicit and explicit have near opposite meanings, so it’s important to remember their difference.
Implicit is indirectly stated or implied.
Explicit is directly stated and spelled out.
If you have any other questions about commonly misused English words, feel free to check out our other posts on affect/effect, principal/principle, and countless others.