User:Alexander L. Davis/Notebook/Error Models and Data Sharing in Hindsight/2012/02/09: Difference between revisions

From OpenWetWare
Jump to navigationJump to search
Line 38: Line 38:


==Corrections==
==Corrections==
* Insert content here...
*Add  the word "theory" to clarify these questions.  
*Need to make the no feedback condition make more sense   
*With data sharing:
1) need to specify the curriculum for them to be able to make judgment 2) Need to specify that the "response" is where the child put the dot and not personal information, etc.  




Revision as of 11:44, 9 February 2012

Error Models and Data Sharing in Hindsight <html><img src="/images/9/94/Report.png" border="0" /></html> Main project page
<html><img src="/images/c/c3/Resultset_previous.png" border="0" /></html>Previous entry<html>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</html>

Entry title

  • Qualitative Interview, participant 9

Comments

  • Rethink Fischhoff's debiasing paper; it seems like this task could be crazy, and we have to make sure it isnt before we conclude that people are crazy;

Unexpected Observations

  • Participant believed that his theory would be 100% true if he expected the child to place the dot in area A EXACTLY 50%of the time and area C EXACTLY 50% of the time; this (A,B,C)->(50, 0, 50) distribution should have very low prior probability though.

New Hypotheses

  • The conditional # and type of revisions is very interesting. This person received no feedback and made no revisions. The new hypothesis proposed was just restatement of the old hypothesis. Other participants have generating new hypotheses of error models, etc.
  • Even highly educated people, such as graduate-level engineers, do not seem to be able to consider multiple sources of evidence and hypotheses at one time. The hypothesis space is very sparse to start, with one initial most plausible model being articulated. We might try to probe to see if they can generate more hypotheses without feedback, with hypothetical feedback, or whether real feedback is required for robust hypothesis generation.

Current Protocol

  • Qualitative interview on Qualtrics; in person; think-aloud and probing questions

Current Materials

New Data

Just one qualitative interview. Other participant was a no-show.

Faults

  • Participant was confused about the P(theory) question. Might want to just use the word "theory" to clarify these questions. People tend to naturally call it a theory, so much bias is not expected to be added.
  • There are confusing questions when the participant does not get feedback: revising hypothesis, etc.; needs to be fixed
  • Need to fix hypothesis revision questions to allow them to skip to end of survey if they dont want to revise
  • With data sharing:

1) need to specify the curriculum for them to be able to make judgment 2) Need to specify that the "response" is where the child put the dot and not personal information, etc.

Corrections

  • Add the word "theory" to clarify these questions.
  • Need to make the no feedback condition make more sense
  • With data sharing:

1) need to specify the curriculum for them to be able to make judgment 2) Need to specify that the "response" is where the child put the dot and not personal information, etc.



Retrieved from "https://openwetware.org/mediawiki/index.php?title=User:Alexander_L._Davis/Notebook/Error_Models_and_Data_Sharing_in_Hindsight/2012/02/09&oldid=582760"