“What is bias?” is a question asked of many people. The answer to the question often depends on who you ask it to. If you are asking a colleague, your friend, an administrator, or even a customer, the answer may be very different than if you were to ask a statistician, engineer, or businessperson.
For our purposes here, we will be discussing what is bias in the context of the human brain sciences. What is bias? In today’s vocabularies, the word “bias” is often used to describe a prejudice, stereotyping, or discrimination. However, the full definition of bias is: “a systematic error made by a human being that tends to favor one belief over another”.
The question as to what is bias turns on how it affects you, the subject of the test, and your ability to understand and recall information. One example of what is bias in the sample size or survivorship bias lens is this: Suppose you were being surveyed about which medical careers are most likely to make you wealthy. You would have to answer questions about all the possible careers (for women) with income, employment history, etc… and then rank them from highest to lowest. The results would then be presented to the research scientist.
Now suppose that you were taking part in a science fair project and you came up with a new idea for a product that solves a problem. You would have to explain your idea and its potential benefits to the research scientists. Your concept might now be regarded as an “example” of what is bias in the sample size or survivorship bias lens.
Another example of what is bias in the sample size or survival bias lens is this: Suppose that there are two groups of people who are being surveyed–a large group and a small group. Within each group of people is someone who is biased toward his/her own cultural competency. This person, the bias subject, will respond in the way that he/she thinks that he/she should, not what the researchers want him/her to do. For instance, if the research indicates that the large group suffers lower IQs than the small group, the bias subject will most likely select the larger career to obtain a higher paying job. This automatic processing is what is considered a form of cultural competency–and it can result in biased results.
Of course, the researchers do try to correct these measurement errors through standard statistical procedures. For instance, if the response variable is measured by using a one-way data normal to fit the normal distribution of the population parameter, the researcher can correct the measurement error by estimating the parameters of the normal distributions to be compared with the measured value. However, the bias subject will still see what is in fact bias in the measurement. Thus, as long as the researcher relies on the “bias” measurement error to adjust the population parameter, the result is an invalid comparison between the two variables. The conclusion drawn from this example is that what is in fact bias in the sample size or level effect of cultural competency can have the same type of invalidations as what is seen in the example above.
It is important to note that there are two different types of biases: conscious and unconscious biases. The former are based on a person’s attitudes towards different groups, whereas the latter are motivated by other factors. Thus, for instance, the gender of the researcher is considered a conscious bias against female employees in one context, while the attitudes of the employee towards sex does not have any effect on whether she is promoted or not. Similarly, unconscious biases such as race, age, or religion cannot be measured, and so are not related to any changes in performance.
The problem of what is implicit bias arises when researchers choose to use explicit measurements and overlook the existence of implicit biases. In other words, when researchers collect and compare data from different sources, they must consider both the implicit and the explicit bias measurement tools. Unfortunately, there are many companies that use implicit measures, but fail to take into account the possibility of implicit bias. Thus, it is important for researchers to remember that what is implicit may appear as implicit in a particular context but actually has the potential to create real-life problems.