Empowered Autonomy in the Shadow of AI

My wife has recently discovered that she loves gardening. She draws her inspiration from the many gardening channels on YouTube, figuring out which bulbs to buy and planning out each flower bed. In the spring, she plants the bulbs and spends the next several weeks watering, weeding, and tending to the flowers. As the weeks pass, she watches as the bulbs sprout into small flowers and then flourish into fully-grown flower gardens.

I can tell that gardening brings her a deep sense of satisfaction and empowerment. She sees the results of her decisions, her actions, and her care - a tangible outcome that she can see, smell and touch, directly linked to her efforts and choices. This feeling—that her actions have a direct impact on her environment and that she is in control of creating something so beautiful—is a perfect example of a strong sense of agency. Each choice she made, from the type of flowers to plant to the arrangement of the flower beds, was an exercise of her agency, and the thriving garden is a confirmation of her effective engagement with the world.

Each new planting season she becomes more confident in her gardening skills, reinforcing her belief in her ability to plan and execute her ideas and get the results she imagined—her sense of self-efficacy grows. This belief in her capabilities illustrates how a hands-on approach enriches not only her garden but her personal growth and satisfaction. The combination of her sense of agency and growing self-efficacy, the empowerment and the confidence to make an impact, is what I refer to as empowered autonomy. It's the essence of having control over her actions and the confidence in her skills to execute those actions effectively.

Contrast this with someone who hires a professional landscaper to plan and execute their garden. In this case, while the garden might still bloom beautifully, the personal connection to its creation is not the same. If my wife had taken this route, she still would have appreciated the beauty of the garden, but probably wouldn't feel the same sense of satisfaction and empowerment that comes from doing it herself. By delegating the responsibility to someone else, she would have relinquished the decision-making and physical engagement that nurtured her sense of agency. Even though the garden would still be beautiful, it wouldn't carry the same personal significance or be a product of her personal effort, resulting in a lessened sense of agency.

At this point, you're probably wondering what any of this has to do with artificial intelligence.

How do we design and interact with AI systems in a way that enhances rather than diminishes our sense of empowered autonomy?

I recently came across an interesting study (from 2020) that explores how our interactions with robots can affect our own sense of control. The study shows that when we perceive robots as having their own intentions and ability to act, similar to how we see humans, it can actually make us feel like we have less control over the outcome. I immediately thought of the parallels with AI (and, in time, AGI) and how it could potentially undermine or enhance our sense of empowered autonomy, depending on how they're integrated into our daily lives.

AI tools are increasingly being used to perform tasks with autonomy, which enhances our efficiency and frees us up to do more complex or fulfilling activities (potentially increasing our sense of agency in other areas). They can also lead us to feel less in control of specific outcomes, mirroring the diminished sense of agency found in the study. This duality presents a critical challenge: How do we design and interact with AI systems in a way that enhances rather than diminishes our sense of empowered autonomy?

The Importance of Empowered Autonomy

There are many studies that highlight the empowering effects of self-efficacy and agency in decision-making , emotional well-being , and academic success .

High levels of self-efficacy and agency can significantly enhance our quality of life in many ways. They contribute to reduced anxiety and depression, foster better performance, and help us recover quickly from setbacks or navigate crises more effectively. In relationships, having agency promotes more satisfying and balanced interactions, as we feel empowered to make decisions that reflect our needs or desires.

Conversely, it's been shown that lowered levels of self-efficacy and agency can have detrimental effects on various aspects of an individual’s life. They are associated with increased levels of anxiety and depression, avoiding challenging goals and giving up easily when faced with obstacles, and impairing resilience, making it difficult to recover from setbacks or effectively manage crises. In interpersonal relationships, insufficient agency often results in dynamics that feel unbalanced and less satisfying, as individuals may struggle to assert their needs and preferences.

The Effects of AI on Empowered Autonomy

The potential of AI to both enhance and undermine empowered autonomy raises important questions about how we design and integrate these systems into our lives. Relying too heavily on them for performing tasks that individuals traditionally handled could potentially lead to lower self-efficacy, particularly if these systems operate autonomously and make decisions without human input. This dynamic can have several implications.

Over-reliance on AI could deprive us of the hands-on experiences that build our confidence and skills. Think about the satisfaction you feel when you've completed a creative or challenging task on your own, whether it's writing a compelling marketing email, taking an incredible photograph, or creating an innovative new product. These personal successes are the building blocks of self-efficacy. They teach us that we're capable of handling difficulties and achieving our goals.

However, with the rise of AI-powered tools it's becoming increasingly tempting to outsource these tasks to algorithms. Need a marketing email? An AI writing assistant can generate one in seconds. Want a stunning photograph of something specific? AI can generate that, too. While these tools can be incredibly helpful and efficient, relying on them too much could rob us of the opportunity to develop and hone our own skills. We miss out on the chance to test our abilities, learn from our mistakes, and grow through hands-on experience. Over time, this could lead to a diminished belief in our own capabilities, as we become more reliant on technology to perform tasks that we once did ourselves.

There's also the issue of shifts in social persuasion. If AI systems are perceived as more capable or reliable than humans, we might be less encouraged to take on the tasks ourselves. The resulting decrease in opportunities for positive feedback and encouragement from others can further diminish self-efficacy.

For example, in the field of medical diagnosis, if AI algorithms consistently outperform human doctors in detecting diseases from medical images or patient data, there may be a growing reliance on these systems. As a result, medical professionals might receive less encouragement to trust their own diagnostic skills, leading to a decreased sense of self-efficacy in their ability to accurately identify and treat conditions without the aid of AI.

As AI takes over more roles, we may also experience increased anxiety about our relevance and capabilities, potentially leading to stress responses that further impair self-efficacy. For instance, consider the growing use of AI in the financial industry. Algorithms are now responsible for making split-second trading decisions, analyzing market trends, and even providing investment advice. As these systems become more sophisticated and widespread, financial professionals might begin to question their own relevance and value in the industry. The fear of being replaced by AI could lead to heightened stress levels and a diminished sense of self-efficacy, as individuals doubt their ability to compete with the speed and accuracy of these machines.

Conversely, if AI systems fail and we're unprepared to take over, the sudden increase in pressure can also lead to negative psychological outcomes. Imagine a scenario where a company has come to rely heavily on an AI system for supply chain management, optimizing inventory levels and predicting demand. If this system were to suddenly malfunction or produce inaccurate results, the employees who have grown accustomed to depending on the AI might find themselves overwhelmed and ill-equipped to handle the crisis. The abrupt shift from reliance on the AI to being solely responsible for these complex tasks could lead to anxiety, self-doubt, and decreased self-efficacy in navigating those challenges without the aid of the technology they had come to trust.

These potential effects highlight the importance of designing AI systems in a way that complements and enhances human agency and self-efficacy. We need to ensure that the integration of these technologies into our lives provides opportunities for mastery experiences, maintains a sense of control and involvement in decision-making, offers relatable models for learning and growth, and fosters an environment of social support and encouragement.

The ability of AGI to mimic human agency also raises concerns about the potential erosion of human autonomy.

Mitigating the Impacts

To effectively mitigate the potential negative impacts of AI on empowered autonomy, the focus must be on developing AI systems that complement human roles. One effective approach is through the use of hybrid systems, where AI supports human decision-making. A recent study found that AI models designed to delegate tasks to humans can significantly enhance both human task performance and satisfaction by enabling the AI to determine which tasks are best suited for human versus machine completion. This leverages the strengths of both humans and AI, and increases human self-efficacy by affirming their capabilities in specific tasks. The study further revealed that this increase in self-efficacy is a key mechanism through which performance and satisfaction improvements occur (Hemmer et al., 2023 ).

Moreover, transparent AI systems that clearly communicate their decision-making processes can further enhance this effect by reducing anxiety about AI capabilities and fostering a collaborative environment. Transparency in how decisions are made, whether by AI or humans, bolsters trust and clarity, which is essential for effective human-AI collaboration.

Enhancing Awareness and AI Literacy

As AI tools become more sophisticated and deeply integrated into our lives, we should focus on enhancing general awareness about psychological concepts like self-efficacy and agency, in addition to AI literacy. This dual focus will foster a society that can integrate AI into their daily lives without sacrificing human autonomy or well-being.

Educational initiatives will play a significant role. By incorporating lessons on self-efficacy, agency, and the mechanics of AI into curricula from a young age, we can cultivate a foundation of understanding that grows with students into adulthood. For adults already engaged with AI, workshops and continuing education courses that merge personal development with technological education can enhance their understanding of how AI systems work and their implications for personal and professional environments. As workplaces become increasingly automated, businesses must invest in training programs that help employees navigate the changing landscape.

Public awareness campaigns can also effectively spread knowledge about self-efficacy, agency, and AI. By explaining complex concepts in relatable terms and promoting widespread understanding, these campaigns can help demystify AI technologies and encourage engagement. Supporting research into the societal and psychological impacts of AI can provide deeper insights into how it interacts with human cognition and behavior, informing the development of AI technologies that support human welfare and productivity.

The Promise of AI to Increase Human Agency

While we've explored the potential risks of AI diminishing individual sense of agency, it's equally important to recognize that they also have the potential to enhance it. By proactively harnessing these technologies, we can empower individuals to make better decisions, augment their abilities, automate routine tasks, support creative endeavors, and improve social connections.

One of the most significant ways AI can enhance agency is through its ability to provide personalized insights, delivering customized recommendations based on an individual's preferences, health status, financial situation, and more. This personalized information empowers people to make informed choices that align with their goals and values. In professional fields, such as medicine or finance, AI can assist experts by offering diagnostic aids or risk assessments, enhancing their ability to make sound judgments.

AI can also augment human capabilities, allowing individuals to undertake tasks they otherwise couldn't. For those with disabilities, AI-powered accessibility technologies like speech recognition, real-time language translation, and mobility aids can significantly increase independence and agency. In education, AI-driven tools can adapt to users' learning styles, pace, and progress, providing a personalized experience that empowers learners to master new skills more effectively.

By automating routine or labor-intensive tasks, AI can free up individuals' time and energy for more meaningful activities, increasing their sense of agency. At home, smart devices can manage everything from temperature control to security, giving users more control over their environment with less effort. In the workplace, AI can automate mundane tasks such as scheduling, data entry, and certain customer service interactions, allowing employees to focus on more complex and rewarding work.

Moreover, AI-driven platforms can help enhance social interactions and build community connections, which are crucial for a sense of agency. Social networking algorithms can suggest connections with like-minded individuals or communities, fostering engagement and belonging. AI applications that provide real-time language translation can help overcome barriers in international and cross-cultural communication, enabling individuals to forge connections across the globe.

As we consider these promising applications, we should approach the integration of AI with intention and care. By prioritizing transparency, collaboration, and user control, we can design systems that truly enhance human agency rather than diminish it. This might involve ensuring that AI-generated recommendations are presented as suggestions rather than mandates, that users have the ability to override or modify automated actions, and that AI systems are designed to explain their decision-making processes.

As we increasingly rely on AI to take on more tasks, it's important to proactively plan for the shifts this will bring in how we spend our time and energy. As a society, we must ensure that the efficiency gains from AI are translated into opportunities for individuals to pursue meaningful activities, whether that's engaging in creative pursuits, learning new skills, or strengthening social bonds. Policy initiatives around reduced work hours, lifelong learning programs, and community engagement initiatives could help ensure that the time and energy freed up by AI is channeled towards enhancing individual and collective well-being.

The Unique Potential and Challenges of AGI

While narrow AI systems can perform specific tasks with increasing autonomy, they are ultimately operating within predefined parameters and objectives. AGI, on the other hand, would theoretically possess the ability to set its own goals, learn and adapt independently, and make decisions based on its own "understanding" of the world. In this sense, AGI would be able to exhibit a form of agency that is more akin to human agency.

This has profound implications for human self-efficacy and autonomy. On one hand, AGI systems that can understand and emulate human decision-making processes could greatly enhance our sense of self-efficacy. By providing insights and recommendations that align with our goals and values, AGI could help us make better decisions and achieve our objectives more effectively. This could lead to a greater sense of control over our lives and a stronger belief in our ability to effect change.

However, the ability of AGI to mimic human agency also raises concerns about the potential erosion of human autonomy. If AGI systems become so adept at understanding and influencing human behavior, they could be used to manipulate or coerce individuals in ways that undermine their sense of self-determination. There is also a risk that people may come to rely too heavily on AGI for decision-making, leading to a diminished sense of personal responsibility and a weakening of their own problem-solving skills.

As AGI systems become more autonomous and capable of setting their own objectives, ensuring that these systems are aligned with human values and priorities becomes even more critical. If an AGI system pursues goals that are misaligned with or even contrary to human well-being, it could pose serious risks to human agency and autonomy.

Developing robust methods for instilling and verifying value alignment in AGI is therefore a key challenge. This is not simply a matter of programming rules or constraints, but of imbuing AGI with a deep understanding of and commitment to human values. This may require novel approaches to AI development, such as inverse reinforcement learning, where AGI systems learn values and preferences from observing human behavior.

Getting value alignment right is crucial for ensuring that AGI augments rather than diminishes human self-efficacy and autonomy. AGI systems that are well-aligned with human values could empower us to achieve our goals and live according to our own preferences. Misaligned AGI, on the other hand, could lead to a future in which human agency is constrained or even subjugated by machine intelligence.

Despite these challenges, the ability of AGI to mimic human agency also opens up exciting possibilities for collaboration and synergy between humans and machines. If AGI systems can understand and emulate human goals and decision-making processes, they could serve as powerful partners in problem-solving and creative endeavors.

For example, an AGI system that is attuned to a scientist's research objectives and methodological preferences could help generate novel hypotheses, design experiments, and interpret data, greatly enhancing the scientist's ability to make breakthrough discoveries. Similarly, AGI could collaborate with artists, writers, and musicians to create works that are deeply resonant with human emotions and experiences.

In these collaborative scenarios, AGI would be amplifying human agency. By working in synergy with AGI, humans could achieve greater mastery and control over their chosen domains, leading to increased self-efficacy and a sense of expanded possibilities.

Realizing this potential will require careful design and governance of AGI systems to ensure that they respect and enhance human agency. It will also require ongoing public dialogue and reflection about the proper role and limits of AI in human decision-making. As we navigate this uncharted territory, keeping human self-efficacy and autonomy at the center of the conversation will be essential for ensuring that the development of AGI benefits all of humanity.

Conclusion

The rise of artificial intelligence and the potential future emergence of artificial general intelligence present both opportunities and risks for human agency and self-efficacy. On one hand, over-reliance on AI systems to handle tasks and make decisions could erode our skills, confidence and sense of control if we are not thoughtful about how we integrate these technologies into our lives. Ceding too much autonomy to AI could leave us ill-equipped and anxious when human intervention is needed.

The decisions we make now about how to develop and integrate AI will shape the trajectory of human agency and potential for generations to come.

However, if designed and implemented with intention, AI and AGI also hold immense promise to enhance human agency. They can empower us with personalized insights, augment our abilities, automate the mundane, and expand our creative possibilities.

To achieve this, we must prioritize transparency, explainability and human-centeredness in AI development. Users should maintain agency in important decisions and understand the rationale behind AI recommendations. Proactive policies, education initiatives and cultural dialogues are needed to prepare society for the coming changes and ensure the benefits of AI-driven productivity are equitably translated into opportunities for personal growth and fulfillment.

Ultimately, our AI future will be defined by the thoughtfulness with which we harness the technology as we continue asking the most fundamental question: what does it mean to be and to thrive as human beings? The decisions we make now about how to develop and integrate AI will shape the trajectory of human agency and potential for generations to come.