Skip to content

MariePittet/Fraudulent_study_participation_detector

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

29 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Fraudulent study participation detector

This Streamlit app detects fraudulent participation to an online study with 3 cognitive tasks (Go-NoGo, Approach-Avoidance Task, 2-Alternatives Forced-Choice). It flags speed-running, excessive inattention, bot completion. Caution: in our case, study participants are informed prior to participation and prior to starting the tasks that such algorithm will be applied and that compensation will be denied for fraudulent behavior.

How to use

  • Open the Streamlit app in your browser
  • Drop participants' csv files: you can drop multiple files per participant, and multiple participants (they will be grouped by ID)
  • Read the fraud report: see if some participants' files are flagged as suspicious and why exactly, inspect trial distributions
  • Chase fraudsters and deny study compensation money

Fraud flagging rules

Rule Threshold Behavior flagged
All tasks speed-running > 15% trials below minimal RTs* Speed-running
GNG high false alarm rate > 30% on NoGo trials Always pressing to rush regardless of stimulus type
GNG high miss rate > 10 % on Go trials Letting the task run while doing something else
2AFC same-side streak > 10 consecutive trials Always chosing the same side to rush
AAT high error rate > 25% incorrect trials Ignoring rules to rush or responding randomly without focusing
AAT same-response streak > 12 consecutive trials Always avoiding or approaching to rush

* Minimal RTs are 250ms for GNG and AAT, and 200ms for 2AFC since preference may induce quicker RTs for highly recognized and liked items than external rules

No data but wanna feel the thrill of catching villains ?

Use the data in fraudulent_data_examples to drop in the app. This is courtesy of our two professional in-lab fraudsters (Dhwani Shah and Hugo Najberg) who tried to bypass filters.

About

Streamlit app to detect fraudulent study participation for compensation money. Flags abnormal response patterns to 3 cognitive tasks

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages