CONSENT TO PARTICIPATE

Describing Attention

Investigators:

Kirsten Ziman, Ph.D., Lecturer, Princeton University

Michael S. Graziano, Ph.D., Professor of Psychology, Princeton University

Key Information:

• Your informed consent is being sought for research. Participation in the research is voluntary. The alternative is to not participate.
• Purpose: The purpose of this research is to train AI’s to describe mental states (e.g., the states of being focused, distracted, and so on). Your data will help train and test these networks.
• Expected duration: 30 minutes-2 hours
• Procedures: You will be asked to complete a behavioral task where you will see images and videos on a screen. You will be asked to describe what you see, either on your own, or working together with an AI.
• Foreseeable risks or discomforts: The risks involved in this behavioral study are minimal and are limited to the risks present during routine computer usage.
• Benefits: In exchange for your participation, you will be paid $20/hour or receive course credit.

Introduction

You have been asked to participate in a study aimed at training AI’s to describe mental states. To pursue this aim, we are collecting descriptions of attention from human participants. These descriptions will help our AI’s learn to describe attention in ways that sound natural. If you agree to participate in this study, you may view multiple videos depicting other agents’ attention moving around the screen (with attention shown as a bright blob, moving over a black background, over an image, or over a movie). Your main job will be to describe the attention you see in each video. You may also be asked to interact with an AI, working together to describe images or videos. In this case, part of your job will be to judge how well the AI works. For example, we might ask you about how friendly, collaborative, and helpful the AI is, and other basic feedback about its functioning. If you decide to participate, your collective response data will be used to train AI’s and to assess AI performance. The alternative is to not participate.

Procedures

In this experiment, you will be asked to sit in a quiet, comfortable location and use your computer’s standard monitor and keyboard to interact with the experiment and submit your responses. The data we collect from you may include any typed or button-press responses, entered on a standard computer keyboard, as well as responses in the form of multiple-choice, drop-down, checklists, etc., entered using a standard computer mouse, and verbal responses, collected via audio recording. We will also ask you for basic demographic information about yourself (e.g. age and gender), to help us track the demographics of the subjects in our study. Collectively, your responses will help us to develop, train, and test AI’s that will be more intuitive and pleasant for users.

In addition to these entered and/or spoken responses, at the end of the experiment we will ask you for responses describing your experience (e.g. parts of the tasks you found especially easy or especially tricky, things you noticed while doing the experiment, etc.). These responses will help to inform our future studies and improve our experiment design. Risks and Inconveniences

The risks involved in this behavioral study are minimal and are limited to the risks present during routine computer usage. For example, prolonged time spent sitting at a computer can cause eye strain or back discomfort. For this reason, study sessions will be limited to 120 minutes or less in length. Additionally, periodic breaks will be built into the study, allowing you to rest and relax (use the restroom, have a drink of water, etc.) and continue, if and when you are comfortable. If you become uncomfortable at any time and wish to stop or withdraw from the study, you may do so immediately with no loss of benefit.

General confidentiality.

The main investigators of this study will not be able to link your name to your data. The only link between your data and your personal identity will be your Prolific ID number. Anonymous data (without your name or Prolific ID number) may be presented at research meetings, published in research journals, and shared in online repositories. Anonymous data may be provided to third parties (such as the creators of data analysis software) for use in connection with product development and marketing activities. Anonymous data may also be entered into artificial intelligence platforms (such as ChatGPT or other similar technologies) by researchers for analysis, hypothesis generation, or other research-related purposes.

Benefits

You will be paid $20/hour for participation in this study if you are recruited as a paid subject, otherwise you will receive course credit. If a session is terminated early due to participant discomfort, software malfunction, or other unforeseeable event, you will nonetheless be paid for the session. In addition to financial or academic remuneration, you may also draw satisfaction from the knowledge that information generated by this study may help improve rapidly emerging technologies. This knowledge may facilitate enhanced collaboration between humans and AI’s and make it easier for society to enjoy the benefits of AI assistants in daily life.

Consent

• I understand that there may be no direct benefit to me from my participation in this study other than the monetary payment or academic course credit described above.
• I understand that my participation will not cost me anything other than the time and effort involved.
• I understand that the information obtained from this study will be confidential and my identity will remain anonymous in any publications or public datasets resulting from this study.
• I understand that by signing this agreement, I do not waive any legal rights or release Princeton University or its agents from liability.
• I understand that this consent is voluntary, that I may withdraw from this study at any time without penalty, and that the alternative is to not participate.
• I understand that if I wish further information regarding my rights as a research subject, I may contact the Institutional Review Board for Human Subjects at Princeton University by telephoning (609) 258-8543 or by sending email to irb@princeton.edu.
• I understand that if I wish further information regarding this research study I may contact Kirsten Ziman by sending an email to kz0108@princeton.edu

I consent that I have read and understood the conditions of consent, and I: