Feeling of Agency (FOA) and Judgment of Agency (JOA) in AI-Enabled Learning Environments

The Feeling of Agency and the Judgment of Agency

Synofzik and colleagues (2008) distinguish between the Feeling of Agency (FOA) and the Judgment of Agency (JOA). While FoA represents the non-conceptual, low-level feeling of being the agent of an action, JoA refers to the conceptual and interpretative judgment of being an agent.

The Feeling of Agency (FOA) is the subjective experience of being in control of our actions and the outcomes they produce. It is the feeling that we are the authors of our own lives and that we have the power to make things happen in the world. It is the background buzz of control over our voluntary actions, even when we're not explicitly thinking about them (Moore 2016). It is a fundamental aspect of human consciousness and identity, as it shapes our sense of self, our motivations, and our interactions with others.

Examples of FOA:

• Reaching out to touch a doorknob: When we reach out to touch a doorknob, we experience a sense of agency over our arm movement. We predict that our arm will move in a certain way, and we receive visual feedback confirming that our prediction is correct—this congruency between our prediction and the actual outcome results in a feeling of agency.
• Playing a video game: When we play a video game, we experience a sense of agency over the character's movements. We press buttons on the controller, and the character moves on the screen accordingly. This congruency between our inputs and the character's outputs results in a feeling of agency over the character.
• Using a self-driving car: When we ride in a self-driving car, we experience a feeling of agency over the car's movement, even though we are not actively controlling it. Through FOA we trust (wisely or otherwise) the AI system to make decisions on our behalf and to keep us safe.
• Using a voice assistant: When we use a voice assistant, such as Siri or Alexa, we experience a feeling of agency over the tasks that we are asking it to perform. This is because the voice assistant responds to our commands and carries out our requests.
• Using a machine translation tool: When we use a machine translation tool, such as Google Translate, we experience a feeling of agency over the translation process. This is because we can choose the languages that we want to translate between and to see the results of the translation immediately.

The Judgment of Agency (JOA) is a higher-level cognitive process that involves making explicit judgments about the causes and consequences of actions. It is a secondary level of agency, built upon the primary level of the feeling of agency (FoA). It is a reflective, intentional, and deliberative process that requires us to consider a variety of factors, such as context, goals, motives, and outcomes. JOA is critical for making sense of our behavior and the behaviors of others, as well as for predicting and controlling the outcomes of our actions.

Examples of JOA

• Watching a video of yourself: When you watch a video of yourself performing an action, such as giving a speech or playing a sport, you are making a judgment of agency. You are deciding whether or not you are the agent of the action in the video.
• Reading about a historical event: When you read about a historical event, such as the American Civil War, you are making a judgment of agency about the people involved in the event. You are deciding who the agents of the event were and who was responsible for its outcome.
• Using a self-driving car: When you ride in a self-driving car and use JOA, you are making a judgment of agency about the car. You are deciding whether or not the car is the agent of its own actions, such as changing lanes or avoiding obstacles.
• Using a voice assistant: When you use a voice assistant and using JOA, you are making a judgment of agency about the voice assistant. You are deciding whether or not the voice assistant is the agent of the actions that it performs, such as setting a timer or playing a song.
• Using a machine translation tool: When you use a machine translation tool and apply JOA, you are making a judgment of agency about the machine translation tool. You are deciding whether or not the machine translation tool is the agent of the translation process.

Mapping FOA and JOA to the Three Paradigms of AI-Enabled Learning Environments

Ouyang and Jiao (2021) group AI and Education into three paradigms, which collectively represent a shift from a teacher-centered approach to a learner-centered approach. AI-enabled learning environments can support both FOA and JOA in learners. For example, AI systems can provide learners with personalized learning experiences that are tailored to their individual needs and abilities. AI systems can also provide learners with immediate feedback and support, which can help them to learn more effectively. As a third example, AI systems can help learners develop their critical thinking and problem-solving skills by providing them with opportunities to collaborate with AI systems on challenging tasks.

AI-directed (learner-as-recipient)

In the AI-directed paradigm, the AI system is in control of the learning process, and the learner is primarily a recipient of the AI's services. The AI system can provide the learner with choices such as the pace of the lesson, the difficulty level, and the types of activities they want to do. AI systems can provide learners with immediate feedback on their progress and structure and support, which can help them learn more effectively. While the AI system can give learners choices about the pace of the lesson, the difficulty level, and the types of activities they want to do, this paradigm can limit their FOA, as they may have limited say in how they learn and no choice in what they learn.

Examples of how AI systems can be designed to support FOA and JOA in AI-directed learning environments

• Give learners choices about the pace of the lesson, the difficulty level, and the types of activities they want to do. This can be done by providing learners with a variety of learning activities to choose from, and by allowing them to control the pace of the lesson by pausing, rewinding, or fast-forwarding.
• Give learners the ability to choose whether or not they want to use certain AI features. For example, learners should be able to choose whether or not they want to receive personalized recommendations, automated feedback, or other AI-powered features.
• Make AI systems more transparent and explainable. AI systems should be able to explain to learners why they are making certain recommendations or providing certain feedback. This can be done by providing learners with information about the system's algorithms, or by allowing learners to see the data that the system is using to make its decisions.
• Give learners the ability to override the system's recommendations and feedback. Learners should be able to choose whether or not they want to follow the system's recommendations, and they should be able to provide feedback to the system on its recommendations and feedback.
• Provide learners with opportunities to reflect on their learning. AI systems can provide learners with opportunities to reflect on their learning by asking them questions about what they have learned, and by providing them with opportunities to write about their learning.
• Provide learners with opportunities to make their own decisions about their learning. AI systems can provide learners with opportunities to make their own decisions about their learning by giving them choices about what to learn about, how to learn it, and how to assess their learning.
• Provide learners with opportunities to learn from their mistakes. AI systems can provide learners with opportunities to learn from their mistakes by giving them feedback on their mistakes, and by allowing them to try again.
• Provide learners with feedback on their FOA and JOA. AI systems can provide learners with feedback on their FOA and JOA by asking them questions about how they feel about their learning and how they make decisions about their learning. AI systems can also help learners to develop strategies for improving their FOA and JOA.

AI-supported (learner-as-collaborator)

In the AI-supported paradigm, the AI system collaborates with the learner, providing support and guidance as needed. It helps learners set and achieve learning goals by providing tailored learning plans and progress-tracking tools. It also supports learners when they are struggling, providing additional resources or breaking down complex tasks. This paradigm gives learners more control over their learning, increasing their FOA. The AI system also helps learners develop their JOA by providing feedback on their choices and actions.

Examples of how AI systems can be designed to support FOA and JOA in AI-supported learning environments

• Give learners choices about how and what they learn. AI systems can provide learners with a variety of learning activities to choose from and allow them to control the pace of their learning. AI systems can also give learners the ability to choose the topics they want to learn about, and the order in which they want to learn them.
• Provide learners with opportunities to reflect on their learning and make their own decisions. AI systems can ask learners questions about their learning, and provide them with opportunities to write about their learning. AI systems can also give learners the ability to make their own decisions about their learning, such as what to learn about, how to learn it, and how to assess their learning.
• Provide learners with feedback on their choices and actions. AI systems can provide learners with feedback on their choices and actions, both positive and negative. This feedback can help learners develop their JOA, as they learn from their mistakes and successes.
• Help learners to develop their critical thinking and problem-solving skills. AI systems can help learners develop their critical thinking and problem-solving skills by providing them with opportunities to collaborate with AI systems on challenging tasks. For example, AI systems can help learners brainstorm solutions to problems and evaluate different solutions.
• Provide personalized learning plans. AI systems can generate personalized learning plans for each learner, based on their individual needs and abilities. This gives learners control over their own learning and helps them to achieve their learning goals.
• Provide progress-tracking tools. AI systems can provide learners with progress-tracking tools that allow them to see how they are doing and track their progress over time. This helps learners to stay motivated and focused on their learning goals.
• Offer learning scaffolds. When learners are struggling, AI systems can provide them with additional resources, such as tutorials, articles, and practice problems.
• Break down complex tasks into smaller ones. AI systems can break down complex tasks into smaller, more manageable steps. This makes complex tasks less daunting and helps learners to achieve them.
• Provide collaborative problem-solving tasks. AI systems can allow learners to collaborate with AI systems on challenging problem-solving tasks. This helps learners to develop their critical thinking and problem-solving skills and also gives them a sense of agency over their learning.

AI-empowered (learner-as-leader)

In the AI-empowered paradigm, the AI system is a tool that the learner uses to achieve their learning goals. The AI system can provide the learner with access to a wide range of learning resources, such as articles, videos, and interactive simulations. The AI system can also help the learner to connect with other learners and experts in their field. The learner is in control of their learning, and they are responsible for making decisions about how and when to use the AI system. This paradigm gives the learner the greatest amount of FOA, but it also requires the learner to have a well-developed JOA.

Examples of how AI systems can be designed to support FOA and JOA in AI-empowered learning environments

• Provide learners with access to a wide range of learning resources. AI systems can provide learners with access to a vast and ever-growing library of learning resources, such as articles, videos, blogs, books, and interactive simulations. This gives learners the ability to choose the resources that are most relevant to their learning needs and interests.
• Help learners to connect with other learners and experts in their field. AI systems can help learners connect with other learners and experts in their field by providing them with access to online communities, forums, and discussion groups. This allows learners to learn from and collaborate with others, which can help them develop their FOA and JOA.
• Provide learners with opportunities to reflect on their learning and make their own decisions. AI systems can provide learners with opportunities to reflect on their learning by asking them questions about what they are learning and why they are learning it. AI systems can also give learners the ability to make their own decisions about their learning, such as what to learn about, how to learn it, and how to assess their learning. This helps learners to develop their JOA, as they learn from their mistakes and successes.
• Help learners to develop their critical thinking and problem-solving skills. AI systems can help learners develop their critical thinking and problem-solving skills by providing them with opportunities to work on challenging problems and projects. AI systems can also provide learners with feedback on their work, which can help them to improve their skills over time. This helps learners to develop their FOA, as they learn that they are capable of solving difficult problems and achieving their goals.
• Provide personalized learning pathways. AI systems can generate personalized learning pathways for each learner, based on their individual needs and interests. This gives learners control over their learning and helps them to achieve their learning goals.
• Provide project-based learning opportunities. AI systems can be used to support project-based learning by providing learners with access to resources, helping them to collaborate with others, and providing feedback on their work. Project-based learning allows learners to work on challenging and meaningful problems, which helps them to develop their FOA and JOA.
• Provide inquiry-based learning opportunities. AI systems can be used to support inquiry-based learning by providing learners with access to information, helping them to develop research skills, and providing feedback on their work. Inquiry-based learning allows learners to explore topics that are of interest to them and to develop their own understanding of the world around them. This helps them to develop their FOA and JOA, as they learn that they are capable of finding answers to their own questions and developing their own knowledge.

The following table presents a comparison of AI-directed (learner-as-recipient), AI-supported (learner-as-collaborator), and AI-empowered (learner-as-leader) learning paradigms.

References

Moore, J. W. (2016). What is the sense of agency and why does it matter? Frontiers in Psychology, 7, 1272. https://doi.org/10.3389/fpsyg.2016.01272

Ouyang, F., & Jiao, P. (2021). Artificial intelligence in education: The three paradigms. Computers & Education: Artificial Intelligence, 2, 100020. https://doi.org/10.1016/j.caeai.2021.100020

Synofzik, M., Vosgerau, G., & Newen, A. (2008). Beyond the comparator model: a multifactorial two-step account of agency. Consciousness and Cognition, 17(1), 219-239. https://doi.org/10.1016/j.concog.2007.03.010