The confirmation bias is a tendency to seek information
to prove, rather than disprove our theories. The problem
arises because often, one piece of false evidence can
completely invalidate the otherwise supporting factors.
Consider a study conducted by Peter Cathcart Wason. In
the study, Wason showed participants a triplet of numbers
(2, 4, 6) and asked them to guess the rule for which the
pattern followed. From that, participants could offer
test triplets to see if their rule held.
From this starting point, most participants picked specific
rules such as goes up by 2 or 1x, 2x,
3x. By only guessing triplets that fit their rule,
they didnt realize the actual rule was any
three ascending numbers. A simple test triplet of
3, 15, 317 would have invalidated their theories.
Known more commonly under hindsight is 20/20
this bias causes people to see past results as appearing
more probable than they did initially. This was demonstrated
in a study by Paul Lazarsfeld in which he gave participants
statements that seemed like common sense. In reality,
the opposite of the statements was true.
This is the tendency to see patterns where none actually
exist. A study conducted by Thomas Gilovich, showed people
were easily misled to think patterns existed in random
sequences. Although this may be a necessary by product
of our ability to detect patterns, it can create problems.
The clustering illusion can result in superstitions and
falling for pseudoscience when patterns seem to emerge
from entirely random events.
The recency effect is the tendency to give more weight
to recent data. Studies have shown participants can more
easily remember information at the end of a list than
from the middle. The existence of this bias makes it important
to gather enough long-term data, so daily ups and
downs dont lead to bad decisions.
Anchoring is a well-known problem with negotiations.
The first person to state a number will usually force
the other person to give a new number based on the first.
Anchoring happens even when the number is completely random.
In one study, participants spun a wheel that either pointed
to 15 or 65. They were then asked the number of countries
in Africa that belonged to the UN. Even though the number
was arbitrary, answers tended to cluster around either
15 or 65.
And you were worried about having too little confidence?
Studies have shown that people tend to grossly overestimate
their abilities and characteristics from where they should.
More than 80% of drivers place themselves in the top 30%.
One study asked participants to answer a difficult question
with a range of values to which they were 95% certain
the actual answer lay. Despite the fact there was no penalty
for extreme uncertainty, less than half of the answers
lay within the original margin.
Fundamental Attribution Error
Mistaking personality and character traits for differences
caused by situations. A classic study demonstrating this
had participants rate speakers who were speaking for or
against Fidel Castro. Even if the participants were told
the position of the speaker was determined by a coin toss,
they rated the attitudes of the speaker as being closer
to the side they were forced to speak on.
Studies have shown that it is difficult to out-think
these cognitive biases. Even when participants in different
studies were warned about bias beforehand, this had little
impact on their ability to see past them.
What an understanding of biases can do is allow you to
design decision making methods and procedures so that
biases can be circumvented. Researchers use double-blind
studies to prevent bias from contaminating results. Making
adjustments to your decision making, problem solving and
learning patterns you can try to reduce their effects.