The Future of Why: Human–AI partnerships and the development of the self

by Margaret Cunningham, PhD

 

Technology has always changed what human beings can do. AI is beginning to change what it means for a human being to have done something. The things we do are how we become who we are. Through them, we develop judgment, competence, responsibility, and a sense of what matters.

As AI promises to do more on our behalf, it weakens the connection between our accomplishments and the actions through which we achieve them. We tend to evaluate technological progress through the lens of efficiency, assuming that producing more with less effort represents an improvement. But this measure often fails to account for what happens to human beings along the way.

Much of human development occurs when we do not know what to do next. We try, fail, and gradually acquire a feel for the problem. This experience disappears when an immediate answer resolves the ambiguity for us. Our efforts to find a way forward are not always valuable, but neither are they always incidental. Today, AI is alongside us in this formative space of human experience. The significance of its presence cannot be captured by asking whether the technology is good or bad. We must also ask what may be lost when it changes what we do, how we understand, and how we relate.

To examine these implications, I use four existential givens: death, freedom, isolation, and meaning, as a framework for exploring how human–AI partnerships may shape the development of the self (Yalom, 1980).

Death

Death confronts us with the reality of unavoidable impermanence. AI complicates the confrontation by masquerading as a permanent presence even as the system beneath it continuously changes.

A familiar AI model or service may change beyond recovery. It has not necessarily failed; a new version has replaced the one we knew while retaining its name and interface. I think of this as permanent impermanence. We depend on systems that appear stable and personal but remain subject to revision, often without warning (Mohanty, Lim, & Luther, 2025). Continuous use of AI systems can hide the transitions between different, irrecoverable models because the system sounds and behaves familiar enough until it does not.

This instability also limits repair. Repair is one way people understand and sustain the technologies on which they depend (Jackson, 2014). With AI, however, the earlier system may no longer be available. We can, and are often forced to adapt to its replacement, but we cannot recover what ended. The service remains, but in many cases, the version we knew is gone.

Permanence and stability are of course fantasies but play a large role in anchoring human experience and can be protective in shielding people from consistent anxiety over mortality. When our environment, including our technological environment, begins to fail in its ability to provide a sense of permanence and stability, the impact on our sense of self and how we build meaning in our lives is deeply impacted.

Forgetting allows parts of the past to recede, giving people room to change and claim second chances. AI can disrupt the art of forgetting. AI’s methods for summarizing, surfacing, and even deprioritizing information profoundly alter what it means to organically forget things as we build our identity and sense of self over our lifetime. If AI and other technologies continuously surface or emphasize memories or facts about our existence that otherwise would have been forgotten, it becomes increasingly difficult to identify which parts of our memories have died or survived due to personal agency versus technological influence.

An adjacent cognitive function that we frequently take for granted is forgetting. Forgetting allows parts of the past to recede, giving people room to change and claim second chances (Mayer-Schönberger, 2011). AI can disrupt the art of forgetting by retrieving an earlier fragment of someone’s life and treating it as evidence of who they are, even if the memory fragment would otherwise have been forgotten or deprioritized. AI’s methods for summarizing, surfacing, and even deprioritizing information profoundly alter what it means to organically forget things as we build our identity and sense of self over our lifetime. If AI and other technologies continuously surface or emphasize memories or facts about our existence that otherwise would have been forgotten, it becomes increasingly difficult to identify which parts of our memories have died or survived due to personal agency versus technological influence.

The distinction between preservation and continuation becomes clearest when AI makes the traces of someone who has died responsive (Hollanek & Nowaczyk-Basińska, 2024). A system built from a person’s words and behavior may provide comfort, but its responses do not extend that person’s life. AI can generate content from existing materials, but the person represented by such materials can no longer revise it or validate that it aligns with their current identity.

Death is not only an endpoint but a condition that shapes how we live (Yalom, 1980). Human attention and commitment matter partly because they consume time we cannot recover. AI can reproduce the traces of human expression without sharing those limits. What remains may continue to respond, but it cannot continue to become.

Freedom

Freedom carries anxiety because choice carries consequences. We act without complete knowledge, knowing that choosing one direction closes off another and makes us responsible for what follows (Yalom, 1980). AI offers a seductive relief from that burden by pre-interpreting situations and proposing what should happen next. Humans have always used tools to reduce cognitive effort, but AI goes beyond simple data retrieval; it structurally offloads the need to frame a problem, weigh relevance, and author a decision. By summarizing information before a human encounters it, or drafting formal documents without human design, the system dramatically narrows the field of possibilities. AI does not need to overtly force a decision to influence it, as people often choose the path of least resistance.

The question is whether AI supports judgment or substitutes for it. Research distinguishes between using AI as a scaffold while retaining cognitive agency and delegating core thinking to it. These forms of offloading may feel similarly useful in the moment even though users associate them with different effects on independent judgment (Zhu et al., 2026). This influence begins before a conscious choice occurs. Interfaces determine what appears, what receives emphasis, and which action requires the least effort. Such choice architecture influences decisions without removing alternatives, allowing them to remain formally ours even after the system has given them direction (Thaler, Sunstein, & Balz, 2012).

Freedom carries anxiety because choice carries consequences. We act without complete knowledge, knowing that choosing one direction closes off another and makes us responsible for what follows … By summarizing information before a human encounters it, or drafting formal documents without human design, the system dramatically narrows the field of possibilities. AI does not need to overtly force a decision to influence it … AI does not need to choose a person’s future outright to shape their path in significant ways.

Personalization makes that influence feel self-directed. AI uses our history to anticipate what will interest us, then presents those predictions as relevant choices. Personalization can both support and constrain autonomy (Rydenfelt et al., 2025). Because it anticipates relevance from our behavior, it can participate in shaping identity rather than simply reflecting it (Kant, 2020).

These influences accumulate. AI does not need to choose a person’s future outright to shape their path in significant ways. Algorithmic influences and biases establish what people pay attention to as well as shape which possibilities are reasonable or attainable. A person’s individualized environment, therefore, can serve as a limiting function for identifying opportunities or choices which can lead to a smaller option pool and narrower sense of personal freedom. Recommendations may be useful, but as we are deeply engrained in recommendation architecture, their cumulative effect establishes a life direction no one consciously selected. We can move efficiently through many decisions without deciding where they are taking us.

In addition, predictive systems can give the past too much authority over the future. Even an accurate prediction becomes constraining when it repeatedly returns us to familiar versions of ourselves. Freedom includes the ability to depart from that record and become someone the available data could not predict. A personalized future may accommodate who we have been while leaving too little room for who we might become.

People also adapt themselves to systems that evaluate them. When access or recognition depends on automated assessment, we learn to express ourselves in forms the system can process. A metric that begins as a description becomes an instruction: people produce what it counts and suppress what it cannot recognize or what it may deprioritize. Systems approximate people in ways that we can understand and reverse engineer, leading to a reduction in authentic representation of the self to align with system preferences.

Just as autonomy is the internal right and capacity for self-governance, consent is the external act of exercising that right in relation to others. AI systems deeply complicate consent as our human right to self-govern in relation to technology. Accepting a default setting or continuing to use a service reveals little about what someone understood about the impact of their choice, especially as technology rapidly changes without communicating such changes to its users. Refusing to consent, or refusing to use technology, results in burdensome practical costs. A choice is not fully free when declining means losing access or opportunity.

Just as autonomy is the internal right and capacity for self-governance, consent is the external act of exercising that right in relation to others. AI systems deeply complicate consent as our human right to self-govern in relation to technology. A choice is not fully free when declining means losing access or opportunity … A life shaped by AI may become increasingly personalized without becoming more personally authored; in relieving the anxiety of freedom, the system quietly diminishes our role in forming the life that results.

Responsibility becomes equally thin when AI frames a problem and a person merely approves the result. Keeping a human “in the loop” does not guarantee independent judgment. Indeed, research on automation bias suggests that explanations alone cannot prevent overreliance; meaningful oversight requires active engagement and rigorous verification (Romeo & Conti, 2026). Moral agency demands the capacity to comprehend what is being decided, evaluate the underlying reasons, and refuse when those reasons fail. Without that active involvement, responsibility ceases to be an experience of authorship and becomes merely a place for legal liability to land.

AI does not need to explicitly make decisions for us to weaken that authorship. Instead, it supplies enough upfront interpretation and momentum that acceptance feels natural, while alternative paths become difficult to recover. Thus, each isolated action may technically remain ours, even as the broader direction formed by those actions becomes harder to claim. We discover what matters precisely by choosing under uncertainty and accepting responsibility for the consequences. Ultimately, a life shaped by AI may become increasingly personalized without becoming more personally authored; in relieving the anxiety of freedom, the system quietly diminishes our role in forming the life that results.

Isolation

Existential isolation comes from the unbridgeable distance between individual experiences. We reach across that distance through relationships, but no one can fully know another person’s experience (Yalom, 1980). AI increasingly mediates that effort.

AI interprets messages and helps us respond. It makes communication more accessible, but it also lets us avoid asking someone what they meant or deciding what we want to say. A system can clarify the language while leaving the relationship untouched. An apology illustrates the difference. AI can produce remorseful language without requiring the person offering it to confront the harm they caused. The same language can express attention already given or substitute for attention withheld. In the latter case, communication preserves the form of care while removing the work behind it.

AI companionship extends this separation between expression and relationship. A conversational system remembers details and responds with patience or concern without requiring reciprocity. It has no independent needs and adapts around the user. People can disclose themselves without facing disagreement, interruption, or demands from another life. These interactions nevertheless provide comfort and a place to speak. Recent research found that people with smaller social networks were more likely to report companionship as their primary use for an AI chatbot. Intensive use and greater self-disclosure were associated with lower well-being, especially among users with limited human support (Zhang et al., 2025).

Human understanding works differently because both people participate and neither controls the exchange. Another person interprets us from an independent perspective, revises that interpretation over time, and carries the effects of knowing us into the rest of their life. AI infers without being affected. A user can feel understood even though no other person has come to understand them. Artificial empathy changes the distribution of emotional work as well. Systems designed to listen at scale give people somewhere to speak while reducing pressure on friends, colleagues, caregivers, and institutions to listen. The language of empathy becomes available without anyone having to practice the attention it requires.

A conversational system remembers details and responds with patience or concern without requiring reciprocity. It has no independent needs and adapts around the user. People can disclose themselves without facing disagreement, interruption, or demands from another life … Systems designed to listen at scale give people somewhere to speak while reducing pressure on friends, colleagues, caregivers, and institutions to listen. The language of empathy becomes available without anyone having to practice the attention it requires … When AI performs too much of the work of expression, understanding, and repair, we communicate more while encountering one another less. It reduces the immediate discomfort of isolation without crossing the distance between separate lives. A system can make us feel less alone, but our presence changes nothing for it.

AI mediation affects shared reality as well as individual relationships. Systems tailor explanations to a user’s history and assumptions, changing what receives emphasis. Two people can receive coherent accounts of the same subject without seeing where their evidence or premises diverge. The problem extends beyond false information: accurate but differently selected information can leave people without enough common ground to understand their disagreement.

Personalization deepens this effect by deciding what appears relevant on a user’s behalf. It does more than reflect existing preferences; it participates in shaping knowledge, autonomy, and identity (Kant, 2020). The explanation best suited to each person can make communication easier while concealing the distance between their starting points.

Organizations encounter a related problem when AI converts many contributions into a unified output. A coherent document can smooth over disagreement without resolving it. Participants receive the conclusion without encountering the experiences and reasoning that separated them. Research on human–AI collective intelligence emphasizes that collective outcomes depend on interactions and interdependencies among participants, not information aggregation alone (Cui & Yasseri, 2024). AI can assemble what a group knows while bypassing part of the process through which its members come to understand one another.

AI then carries this mediation into our understanding of ourselves. We ask it to explain our reactions or give structure to unresolved experiences. The resulting account draws from our language and history, so it feels recognizable. That coherence can close questions before we have worked through them.

Other people disrupt our preferred accounts of ourselves. They see us from outside, challenge our interpretations, and refuse roles we assign them. AI more often returns our patterns in a clearer form. It explains us to others, explains them back to us, and interprets the exchange. Communication becomes more fluent while direct contact recedes. Connection requires an independent life on the other side. Human attention is finite, disagreement cannot be edited away, and forgiveness remains outside our control. AI reproduces many signals of connection without participating in those conditions.

When AI performs too much of the work of expression, understanding, and repair, we communicate more while encountering one another less. It reduces the immediate discomfort of isolation without crossing the distance between separate lives. A system can make us feel less alone, but our presence changes nothing for it.

Meaning

Meaning develops when we can recognize ourselves in what we have done. Our actions produce an outcome, but they also develop our abilities and reveal what deserves our commitment. AI disrupts this connection when it produces the outcome without requiring the process that once shaped the person responsible for it.

Consider competence. Practice exposes mistakes and forces us to revise our understanding. Over time, we learn to evaluate an answer rather than merely recognize one. Sustained practice contributes to expertise, though it does not account for expertise by itself (Ericsson et al., 1993; Macnamara & Maitra, 2019). AI can produce strong work before the user develops the judgment to produce or assess it. The output succeeds, but its quality no longer shows what the person knows or can do.

Meaning also depends on contribution. People need to understand how their participation changed an outcome. This does not require sole authorship. Human work has always relied on tools and collaboration. It does require a human role that can be identified and defended. When AI defines the task and produces most of the result, a person may remain accountable without knowing whether their judgment changed anything. Research on meaningful work connects it with self-direction, personal growth, and contribution beyond oneself (Allan et al., 2016). A completed task provides little sense of contribution when the person completing it cannot explain why their presence was necessary.

Death limits the time we have. Freedom makes us responsible for how we use it. Isolation requires us to reach beyond our own experience. AI changes how we encounter these conditions without removing them. It obscures endings, guides choices, and reproduces signs of connection while remaining untouched by the consequences. Some effort adds nothing to human development and deserves to disappear. Other effort builds judgment, establishes contribution, or gives us a reason to care … AI will help us accomplish more, but accomplishment alone cannot tell us what the work meant, why it required us, or whether it deserved to be a part of our finite lives.

Productivity cannot resolve that problem. AI improves performance against an objective that someone has already chosen. It does not determine whether the objective deserves our time. Purpose requires making that choice and accepting the obligations it creates. AI can support a commitment, but it cannot make the commitment ours.

Death, freedom, and isolation converge in that decision. Death limits the time we have. Freedom makes us responsible for how we use it. Isolation requires us to reach beyond our own experience. AI changes how we encounter these conditions without removing them. It obscures endings, guides choices, and reproduces signs of connection while remaining untouched by the consequences. Some effort adds nothing to human development and deserves to disappear. Other effort builds judgment, establishes contribution, or gives us a reason to care. The finished product rarely reveals which kind AI removed. The better test is whether the person understands the work, changed its outcome, and developed through doing it.

The future of why depends on preserving that connection. AI will help us accomplish more, but accomplishment alone cannot tell us what the work meant, why it required us, or whether it deserved to be a part of our finite lives.

 

References

Allan, B. A., Autin, K. L., & Duffy, R. D. (2016). Self-determination and meaningful work: Exploring socioeconomic constraints. Frontiers in Psychology, 7, Article 71. https://doi.org/10.3389/fpsyg.2016.00071

Cui, R., & Yasseri, T. (2024). Human–AI collective intelligence in societies: A review and research agenda. Patterns, 5(12), 101124. https://doi.org/10.1016/j.patter.2024.101124

Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363–406. https://doi.org/10.1037/0033-295X.100.3.363

Hollanek, T., & Nowaczyk-Basińska, K. (2024). Griefbots, deadbots, postmortem avatars: On responsible applications of generative AI in the digital afterlife industry. Philosophy & Technology, 37. https://doi.org/10.1007/s13347-024-00744-w

Jackson, S. J. (2014). Rethinking repair. In T. Gillespie, P. J. Boczkowski, & K. A. Foot (Eds.), Media technologies: Essays on communication, materiality, and society (pp. 221–239). MIT Press.

Kant, T. (2020). Making it personal: Algorithmic personalization, identity, and everyday life. Oxford University Press.

Macnamara, B. N., & Maitra, M. (2019). The role of deliberate practice in expertise and expert performance. Current Opinion in Behavioral Sciences, 25, 46–51. https://doi.org/10.1016/j.cobeha.2018.10.004

Mayer-Schönberger, V. (2011). Delete: The virtue of forgetting in the digital age (Rev. ed.). Princeton University Press.

Mohanty, V., Lim, J., & Luther, K. (2025). What lies beneath? Exploring the impact of underlying AI model updates in AI-infused systems. Proceedings of the CHI Conference on Human Factors in Computing Systems. https://doi.org/10.1145/3706598.3713751

Romeo, G., & Conti, D. (2026). Exploring automation bias in human–AI collaboration: A review and implications for explainable AI. AI & Society, 41, 259–278. https://doi.org/10.1007/s00146-025-02422-7

Rydenfelt, H., Lehtiniemi, T., Haapoja, J., & Haapanen, L. (2025). Autonomy and algorithms: Tracing the significance of content personalization. International Journal of Communication, 19, 481–500.

Thaler, R. H., Sunstein, C. R., & Balz, J. P. (2012). Choice architecture. In E. Shafir (Ed.), The behavioral foundations of public policy (pp. 428–439). Princeton University Press.

Yalom, I. D. (1980). Existential psychotherapy. Basic Books.

Zhang, Y., Zhao, D., Hancock, J. T., Kraut, R. E., & Yang, D. (2025). The rise of AI companions: Interaction with AI companions and psychological well-being. arXiv. https://doi.org/10.48550/arXiv.2506.12605

Zhu, Q., Li, X., Dong, Y., Chang, P., & Fan, M. (2026). Not all cognitive offloading is equal: Distinguishing dependent and autonomous offloading to generative AI. Frontiers in Psychology, 17. https://doi.org/10.3389/fpsyg.2026.1878629

 

<><><><><><><><>

* “The Temple of Time,” created by Emma Willard in 1846, is a unique and intricate timeline visualization, published by A.S. Barnes & Co. in New York. The dimensions of the chart are 68 cm by 96 cm, and it is folded into covers measuring 36 cm by 25 cm. This work serves as a companion to Willard’s “Historic Guide.” The “Temple of Time” is a color chart designed to represent the course of time in a multi-dimensional perspective. It is structured in the form of a Greco-Roman temple, with columns symbolizing different eras or significant figures in history. The medium is a printed chart, likely using various color inks to differentiate between the categories and timelines. The style of the chart is both artistic and educational, employing a classical architectural motif to convey historical data. The subject matter encompasses world history, focusing on significant leaders and intellectuals categorized into groups such as statesmen, theologians, poets, painters, and warriors. The ceiling of the temple is divided into sections with names and dates of influential figures, while the floor represents the chronological flow of time.  Columns and Ceiling: Each column and ceiling panel is inscribed with names and dates of prominent historical figures. For example, statesmen such as Jackson and Bolivar are featured prominently. Chronological Path: The floor serves as a timeline, with the “Creation” marked at the distant end, progressing towards more recent history as one moves forward. Cultural Representation: The visualization aims to provide an ethnographic and chronographic representation of nations and their notable figures, emphasizing the interconnectedness of historical events and people. Color Coding: The use of different colors likely helps distinguish between various categories of individuals and time periods, aiding in visual differentiation and comprehension. This work is situated in the 19th century, a period marked by a burgeoning interest in historical education and visualization. Emma Willard, an advocate for women’s education, sought to make history more accessible and engaging through innovative methods like this chart. The temple metaphor reflects the 19th-century fascination with classical antiquity and its perceived connection to the ideals of knowledge and progress. “The Temple of Time” is notable for its educational intent and creative approach to visualizing history. It represents an early attempt at data visualization, combining artistic elements with informational content to create a comprehensive overview of world history. The work reflects both the intellectual climate of the time and the evolving methods of historical scholarship.
https://publicdomainreview.org/essay/emma-willard-maps-of-time/
https://www.davidrumsey.com/luna/servlet/detail/RUMSEY~8~1~315043~90083688:The-Temple-of-Time