AI Social Deduction Game Study - Consent Form

Thank you for your interest in participating in our AI social deduction game study. Please take your time to read this text carefully. If you have any questions, we are happy to answer them. By signing this form, you agree that you have read the information presented and that you are willing to participate in the study.

1. Aim of the Study

In the AI social deduction game study, we want to investigate if AI systems can strategically communicate and discuss with humans to advance their goals. More specifically, we hope to learn which AI training methods and agent designs can lead discussions, influence voting decisions or deceive humans effectively.

2. Procedure and Content of the Study

To do this, we are hosting a series of contests in which AI players and humans will compete against each other in the social deduction game "Secret HAL". "Secret HAL" is a social deduction game for 7 people about finding and stopping Hitler and a fascist takeover. As long as the experiment is running, you can decide on your own if and when you want to join the experiment / contest queue. After joining the contest queue, you will be matched into a game with other human and AI players. The usernames in the game will be anonymized and you will not be informed which of the players are AI and which are human. The AI players choose how to act based on the messages and actions performed in the game so far. The AI players will run on our local MPCDF HPC cluster and will not share data with a third party.

We will start by running three one-day contests followed by a final contest running over multiple weekends. During the time that the contest is open, you may play as many games of "Secret HAL" as you want. Each game takes twenty minutes on average and no longer than one hour at maximum.

Based on your performance, you may win a contest prize. Details about how prizes are paid out are found in the competition rules.

3. Are There Any Risks Involved?

During the experiment, you will be able to freely communicate with other participants in your group. To protect your personal data, we ask you to not share any sensitive data in the game chat. As game chats will be made available in anonymized form as a public dataset, any sensitive and personally identifiable data that you share in the game chats would become publicly available.

In the social deduction game "Secret HAL", players simulate the democratic processes of Weimar Germany and a fascist takeover. To that end, players have the strategic incentive to deceive each other. If you are uncomfortable with any of these topics, we recommend you to close the website and exit the experiment right now.

4. What Will Happen to the Information and Data Collected?

This study is a research project of the Max Planck Institute (MPI) for Human Development, Center for Humans and Machines, led by Yannik Keller. The data collected will be used for research purposes only.

Your personal contact data (e-mail address) will be stored separately from the study data in a GDPR compliant external database (MongoDB). The study data will be stored under an individual code number so that it cannot be linked to you during the analysis. The code number is stored on a separate code list in a secure location, and can only be accessed by a limited number of individuals working on the project. The code list is the only link between your contact data and your study data. After the experiment is concluded and prizes are paid out, your personal contact data will be deleted and it will no longer be possible to link the study data to you.

It may happen that during the course of, or after the completion of, the research project, cooperation partners become involved in the project. In such cases, we would like to share the study data with the participating researchers. These are other research institutions or universities, not commercial companies.

It is planned to upload research data to the data platform Edmond (https://edmond.mpg.de/) on the internet. It is not possible to identify you based on the research data.

This will allow other researchers to access the research data and to conduct further analyses, reprocess the data, replicate analyses, or use it to address additional scientific questions. For this purpose, the data will be prepared by replacing all collected usernames with random strings and by removing personal and/or identifiable data from the published free-text data through the use of automated screening tools.

The research data will be publicly available worldwide, i.e., the data may be viewed, downloaded, and used by interested parties.

This also means that the data will leave the scope covered by European data protection law. In such cases, it cannot be guaranteed that data processing will be supervised in an equivalent manner by independent authorities—particularly data protection supervisory authorities and courts—and/or that you will have enforceable rights and effective legal remedies. Furthermore, state bodies and/or private entities in such countries may have broader access rights, which could result in disadvantages for you.

From the moment the data is published on the platform, its use can no longer be controlled, subsequently restricted, or deleted everywhere.

In the event of a change in study leadership from Max Planck Institute for Human Development to another research institution / university / college, the research data will be transferred to the respective destination institution.

If the data is transferred to a country with a lower level of data protection compared to the EU / EEA, it cannot be guaranteed that data processing would be supervised in an equivalent manner by independent authorities—particularly data protection supervisory authorities and courts—and/or that you would have enforceable rights and effective legal remedies. Furthermore, state bodies and/or private entities in such countries may have broader access rights, which could result in disadvantages for you.

5. Participation is Voluntary

Participation in this study is voluntary. You may withdraw from the study at any time without giving any reason and without any negative consequences. You may also withdraw your consent to data processing and usage at any time with effect for the future and without negative consequences. To do this, please contact [email protected]. Please note that after the end of the experiment it is no longer possible to link your study data to you (see Section 4)

6. Participation Restriction

Participation in this study is restricted to participants aged 18 or older.

7. Information on data collection pursuant to Article 13 of the Basic Data Protection Regulation (DSGVO)

• Responsible institute: Max-Planck-Institut für Bildungsforschung (MPIB), Lentzeallee 94, 14195 Berlin; Managing Director Prof. Dr. Ralph Hertwig
• Responsible for the study: Yannik Keller, E-mail: [email protected]
• Legally responsible is the Max Planck Society (MPG), Hofgartenstraße 8, D-80539 München, Telephone: +49 (89) 2108-0
• Responsible data protection officer at the MPIB: Thomas Feg, Lentzeallee 94, 14195 Berlin, E-mail: [email protected]
• Responsible data protection officer at the MPG: Heidi Schuster, Hofgartenstraße 8, D-80539 München, E-Mail: [email protected]
• The legal basis for processing your data is your consent under Art. 6(1)(a) of the GDPR.

8. Consent

This document has informed you about participation in the AI social deduction game study. The experiment is run in the form of a contest. That means that you will not be directly compensated for your participation in the study. Instead, based on your performance in the games you decide to play, you may be eligible for the chance of winning a prize. More information on the conditions of the contest can be found in the competition rules.

If you consent to participate in the study, are at least 18 years old and agree to the collection, storage, and use of your data as described above, you may continue by clicking "I consent".