Lab 3: Psychological phenomena in text and picture generations

Goal. The goal of this lab is to study psychological phenomena in AI based text and picture generation software.

Preparation. You do not need to prepare before the lab. But think about what psychological phenomena and model that you want to study. Bring a computer to the lab.

Report. You need to write to complete a report no later the week after the lab.

Below you find the steps that you need to do

1. Create work groups. Divide yourself into groups of 2 to 3 students. List the names of the group in the Google Doc Spreadsheet.

2. Choice text or picture generation to study. Choose so that there are equally many groups text as picture generation groups.  Write your choice (picture/text) in the Google Doc Spreadsheet.

3. Play around with with text or picture generation software to get familiar with possibilities and limitations. 

4. Choice a psychological phenomena to study. Do not choose phenomena that any other groups has selected. Write your choice or model (e.g., ChatGPT3, stablediffission) and psychological phenomena (e.g., TOM) in the Course Spreadsheet .

  • Stereotypes (i.e., gender, race, age, occupation, etc). 
  • In and out group evaluation biases (i.e., are your own group valued higher than the the other group)
  • Theory of mind (TOM); i.e. does the text model understand other people's mind.  
  • Chinese room, i.e., ask questions and ask people if the response was made by a human or AI. 
  • Come up with our own psychological phenomena to study! 

5. Search the literature for at least one articles on the psychological phenomena and at least one on either text or picture generated models. Read the articles. List the articles in the Google docs document.

6. Hypothesis and methods. Come up with hypothesis on how your chosen AI model behavior on your chosen psychological phenomena (e.g., "AI generated picture of occupations shows predictable gender stereotypical behavior). Define a method for testing the hypothesis (For example Let your AI model "Draw a picture of doctor/nurse/CEO/hairdresser" and have humans rate weather they show gender stereotypical behavior).  Write up the hypothesis and method in a Google Doc document and share the link in the Google docs document.

7. Approval. Have your teach approve the method prior to the data collection

8. Data collection using your written up methods. Follow your method and collect the data using the AI model and conduct a quick data collection on five to ten human participants (if is is required by your method).

9. Analysis your data. Make a statistical test to show whether your hypothesis was supported by your data. 

9. Write a short paper on your study. The paper should be 3 to 5 pages long. Write the paper in the Google Docs and share a link in the Google docs document. The paper should be completed at noon the day before the presentation so that opponent groups have time to critically read it. The paper should include the the following headings. 

a. Title page including: Title of the study and group members

b. A conceptual overview of how text or picture generation software works 

c. A short literature review of the chosen psychological phenomena

d. Hypothesis

e. Method

f. Results

g. Discussion

h. Conclusion

10. Make a written review of another groups written study. Group 1 will be reviewer on group 2, Group 2 will be reviewers on group 3, etc. The last groups will be reviewers on the first group.  

11. Correct your paper based on the opposition of another group. Group 1 will be opponent on group 2, Group 2 will be opponents on group 3, etc. The last groups will be opponent on the first group.  

12. Presentation. At the end of the lab, present what you have dones so far. If you are not dones with your data collection or analysis, just present what you have done. 

 

The aims of the seminar are:

  • Understanding the relationships between human and artificial behavior'
  • Learning methods to study AI generated behavior and comparing it with human behavior
  • Train setting up research questions, hypothesis and to methods to test them
  • Writing up a scientific report

 

Things to think about

  • Remember that the other groups will not have read what you have read. Therefore, explain things on a very basic level.
  • There is a limited amount of time. Therefore, use it wisely and focus on the most relevant things.
  • The seminar is mandatory (presence and active participation).
  • The seminar is not a test, but show up prepared!
  • It is possible to fail the seminar.
  • Stick to the subject (AI and psychology).
  • We expect this to stimulate discussion of AI and human behavior!