Lately we hear people and experts talking about random samples. We want to know how effective masks are or how many people had positive results from the treatment. So we start with a smaller number of test cases or people in a poll. Are we actually getting a reasonably accurate assessment?
In critical thinking, sampling or polling can be subjected to bias in a number of ways. One would be the overemphasis on consistent evidence. It is common to use a smaller group for our testing to determine a percentage to use as an estimate for the larger group of the whole. We see this lately when people talk about polling data for elections or virus treatments. It is too time consuming to check everyone in those cases and need to make judgments using these small groups as a substitute. But we also need to recognize their fallibility.
Let’s say we asked an employee to look into how many people ordered alcohol with their meal at the restaurant for dinner. After checking the first 10 customers, he found the task time consuming and pretty much the results were all similar. So he decided to stop after the first 10 and estimate the results from there. He reported that only 1 out every 10 customers were ordering alcohol. He was wrong. He grumbled about wasting time as he continued to find similar results in the next few customer orders. But then as he looked at checks that were from later in the evening, the amount of alcohol ordered increased significantly. What the lazy employee did not realize was that the early customers coming in for dinner were mainly senior citizens and retirees, and the younger customers were the ones mainly ordering wine and drinks with their dinner.
We can see similar biases playing on our assessments by watching early election returns. Judging by results in the first few hours voting we can see from experience that the percentages can change as the day goes on. The reason we cannot accurately predict or assess is because our sampling is not fully representative of the whole. If one starts making judgements on initial results because there are trends that look to be showing consistency, it is tempting to put stock in that trend's accuracy.
To avoid this bias, a critical thinker should understand that there might be some built in trends in the sampling group if it is not truly random. As we are looking to solve a problem or provide an explanation for a situation, we might find that pieces of information that are consistent should all be treated as one piece of evidence instead of multiple pieces due to so much being unknown because the sampling group is not large enough or representative enough of the overall situation.
Please give me a “like” and share with others. Thank you for reading.
Opmerkingen