- Confirmation Bias:
- the seeking or interpreting of evidence in ways that are partial to existing beliefs, expectations, or hypotheses at hand
- often happens when making a decision based on limited information
- assertion: will affect how we seek and interpret information
- stereotyping can result in confirmation bias
- [Bill] in Bayesian Framework, distinct
- interpretation == bias on priors
- seeking == bias on outcomes
- Positive Tests Strategies
- not always a bad thing
- Wason Rule Discovery Task
- feedback is deterministically accurate, rather than probabilistic
- doesn’t tell you where your rule is wrong
- not reflective of feedback in real life
- restricts generalization to other tasks
- essentially same as CS students debugging
- in debugging, students tend not to understand how to make a program give informative information about a problem
- students also do not understand how to logically eliminate a potential source of error
- need to understand the 5 Venn Diagrams
- important is testing the cases most likely to prove you wrong
- Implication for visualization: takeaway
- consider in design process: what we present to people?
- help users say what they need, but also what they don’t need.
- possible use in testing of visualizations?
Klayman assumes binaries: right and wrong, a decision with cost and a decision without cost, and a fixed amount of cost
people aren’t binary, but also can’t appear to comprehend non-binary probability…
If you know what the decision is, why visualize? Just report the answer.
- Often in visualization, the user has a large and vague concept of the design: i.e. a larger hypothesis than the eventual target set
- visualization is needed specifically for the process of hypothesis development
- visualization helps transform ill-posed questions into well-formed ones, which hopefully can eventually be solved using algorithms
Ill-formed or partially-formed case: DEQ, political decision aspect