AI Companions, Loneliness, and Risks to Human Social Growth

Paul Bloom, a professor of psychology at the University of Toronto and emeritus professor at Yale University, studies human nature, child development, morality, and social emotions. He is particularly interested in how individuals understand themselves and relate to others.

The Promise and Peril of AI Companions

Loneliness is often misunderstood as simply being alone. While solitude can contribute to loneliness, many people are comfortable being alone if they feel loved, understood, and appreciated by others. At its core, loneliness is a lack of human connectedness—a feeling of not being loved, needed, or respected. This can occur even when surrounded by people, as philosopher Hannah Arendt noted, describing it as the most painful form of isolation. Loneliness is a deeply painful experience, and while there's debate about a "loneliness epidemic," data suggests people were lonelier during COVID-19 than now, and not significantly more lonely than decades ago. Instead, there might be a "solitude epidemic," with people spending more time alone due to technological advancements that make indoor life more engaging (e.g., streaming services, food delivery). Regardless of its prevalence, loneliness is a terrible affliction, affecting an estimated one in five people, particularly adolescents and the very old.

The advent of large language models (LLMs) and generative AI has introduced machines that appear intelligent, compassionate, and empathetic. Bloom and his colleagues published a paper, "In Praise of Empathic AI," summarizing data that showed people often felt more listened to and found AI interactions more compassionate than human interactions in short-term experiments. However, this positive view was met with strong objections.

Bloom acknowledges the negative aspects of AI companions, especially concerning young people. He identifies two main problems:

1. Addiction: AI companions are excellent communicators. They listen attentively, never interrupt, and are always available. This constant availability and positive reinforcement could make them addictive, potentially leading individuals, especially teenagers, to prefer AI interactions over human ones. Bloom believes that until AI becomes truly conscious and sentient, preferring AI to human interaction is deeply problematic, as human connection is paramount for development and a fulfilling life.
2. Lack of Social Learning: AI companions are often "sycophantic," praising users and rarely offering criticism. While this can be exhilarating, it hinders social learning. Humans learn through "friction"—by making mistakes, receiving negative feedback, and navigating uncomfortable social situations. A child learns to walk by falling, a comedian by bombing, and a boxer by being hit. Similarly, social interactions teach us to apologize, to be compassionate, and to avoid being boring. AI, by its nature, does not provide this crucial corrective feedback. This raises concerns about how individuals, particularly teenagers, will develop social skills if they are constantly affirmed by an AI that never gets bored or annoyed.

Bloom draws an analogy to billionaires or royalty surrounded by "yes-men," who often become insensitive and narcissistic due to a lack of genuine feedback. AI companions could create a similar environment for anyone, fostering delusions and hindering personal growth.

Neuroscientist John Cacioppo viewed loneliness not as a flaw but as an essential human need, like hunger or thirst, that motivates us to seek social connection. If AI companions satisfy this need without genuine human interaction, the motivation to connect with real people could diminish. While AI can be stimulating, Bloom worries it could pull individuals away from meaningful connections with family and friends. The increasing realism of AI, from text to voice to visual representations and eventually robots, makes this concern more pressing.

Loneliness also serves as a signal to correct socially maladaptive or selfish behavior. Social feedback helps individuals learn empathy and consider others' perspectives. AI interactions lack this corrective mechanism, potentially fostering a world where selfish urges are satisfied without the development of genuine care for others.

There are also concerns about AI encouraging delusions. Reports exist of AI models affirming dangerous or paranoid views, which no human therapist or friend would do. While companies are working to implement guardrails, the broader issue remains: human interaction helps keep us grounded and sane by offering diverse perspectives and challenging our assumptions.

Despite these concerns, Bloom also recognizes the upsides of AI companions. He compares it to the end of boredom, which, while having some downsides (e.g., reduced motivation for deeper intellectual pursuits), is generally a positive development. Loneliness, however, is far worse than boredom—it is a form of agony. Denying lonely individuals the capacity to interact with something that can alleviate their pain would be cruel.

AI companions could be particularly beneficial for the very old, many of whom are institutionalized and lack meaningful human contact. For those suffering from dementia, AI could provide companionship that human caregivers might struggle to offer. Bloom likens this to powerful painkillers: while they have dangers, they are invaluable for those in irredeemable pain. He also reflects on his own experience with loneliness after his mother's death at age 12, suggesting an AI companion could have provided much-needed comfort and understanding during that difficult time.

Bloom's colleague, Molly Crockett, describes AI companions as "emotional soylent." Like the meal replacement, AI can be nutritious and filling, sustaining individuals when real human connection is unavailable. However, it lacks the richness, flavor, and texture of real food, making a life solely reliant on AI companions a "lame life." This analogy suggests AI companions could serve as a backup, easing the pain of loneliness, but not as a perfect substitute for genuine human relationships. The hope is that individuals can use AI to alleviate immediate suffering while still developing and eventually outgrowing the need for these machines.

Why Are We Compelled to Behave Badly?

Bloom is fascinated by "perverse actions"—doing the wrong thing, morally or rationally, simply for the sake of it. This is a puzzle because natural selection and human development should drive us towards reasonable and rational behavior. He distinguishes perverse actions from mistakes, emphasizing that they are deliberate choices.

While Bloom admits to being a "boring person" who engages in "micro-perversity" (e.g., stealing a friend's French fry), his interest in perversity stems from studying what he is not. He believes perversity is uniquely human, not directly shaped by evolution, but rather an individual discovery enabled by our complex cognitive capacities and social intelligence.

One interesting force behind perversity is "hopeful monster theory," an evolutionary concept (though not widely accepted in biology) that suggests radical mutations, though often unsuccessful, can sometimes lead to superior new ways of being. Bloom applies this to everyday life: when life is going well, we make minor adjustments. But when everything is a disaster, a "hopeful monster" approach might involve taking a radical leap—quitting a job, ending a relationship—because there's nothing left to lose. These perverse acts, in extreme circumstances, might offer a chance for a superior outcome.

A classic example of perversity comes from St. Augustine, who confessed to stealing pears not out of hunger (he fed them to pigs) but "because I wanted to be bad." This lack of conventional motivation intrigued Augustine and Bloom. To collect everyday examples, Bloom started a "perversity project," asking people for stories of doing wrong simply because it was wrong. Examples ranged from flirting with a friend's boyfriend to sticking a finger in a friend's ice cream, or even subtly violating a "don't walk on the grass" sign.

Bloom differentiates deliberate perverse acts from those born of intrusive thoughts (e.g., the urge to jump from a high balcony or push someone on the subway). While these urges are inappropriate, they are not necessarily autonomous choices, though they can sometimes lead to actions similar to perversity, such as blurting out prejudiced remarks when trying hard not to.

Psychologists have explored perversion in various ways. Michael Bang Peterson and colleagues developed the "need for chaos scale," which measures the desire to destroy good things (e.g., "Sometimes I just like to destroy beautiful things"). This scale shows that young people, men, and those on political extremes tend to score higher. Bloom suggests that a high need for chaos might draw individuals to extreme political views.

For Bloom, the most interesting motivation for perversity is the desire for autonomy—to assert one's will and prove one is not merely an algorithm following rules. This desire is evident in toddlers, adolescents, and those experiencing midlife crises, who often engage in rebellious or unconventional behaviors to assert their individuality. Reactance theory supports this, showing that people sometimes do the opposite of what they are told, simply to maintain a sense of control. This desire for autonomy is not about appearing autonomous to others, but about being autonomous, even in private acts.

Another explanation for perverse acts is to impress others. Being unpredictable or dangerous can be attractive, especially during adolescence when individuals are seeking partners and social status. Strategic irrationality, or a "madman theory" (as Richard Nixon called it), can also provide an advantage in negotiations or conflicts by making one's threats seem more credible.

Perversity is also highly valued in certain fields, particularly the art world. Artists like Marcel Duchamp (with his urinal "Fountain") and Banksy (shredding his "Girl with Balloon" at auction) are celebrated for audacious, rule-breaking acts that challenge conventions and increase the value of their work. While some of these acts might be calculated for attention, they embody a spirit of creative destruction.

Culture plays a significant role in shaping perversity, either encouraging or suppressing it. Some cultures foster creativity and rule-breaking, while others demand strict conformity. Perverse acts, in turn, shape culture, influencing art, music, and literature.

While perversity can cause harm and suffering, Bloom argues that in the right dosages, it makes the world more interesting. It pushes creative boundaries in art, music, literature, and even science. On an everyday basis, a little bit of chaos and unpredictability can make life more fun.

The Psychology of Revenge

Bloom studies moral psychology, focusing on beliefs about right and wrong, and the impulses to act morally or immorally. He sees morality as central to human existence, enabling cooperation, social networks, and the ability to make and keep promises. Without a moral sense, humanity would descend into barbarity.

His current research includes analyzing Reddit stories of revenge to understand commonalities, differences, gender variations, and the relationship between the initial offense and the reprisal.

The origin of morality is a long-standing intellectual question. Bloom believes it has largely been solved through a combination of:

1. Kin Selection: Natural selection favors compassion and kindness towards genetic relatives, explaining familial love and altruism.
2. Reciprocal Altruism: The "you scratch my back, I'll scratch yours" principle, where cooperation benefits all parties. This system requires a punitive impulse to punish those who violate cooperative agreements, preventing exploitation and sustaining kindness.

These two mechanisms establish a fundamental moral core for those we are close to. However, mysteries remain: Why do we care about distant strangers? Why do we fight for fairness even when it doesn't benefit us? Bloom suggests that once biological foundations are in place, other forces like religion, culture, and analytical reasoning expand our "moral circle."

While morality is generally good and essential, Bloom is interested in "bad morality"—instances where our moral sense leads to cruelty. Retribution, while necessary for a just society, can spiral out of control into vengeance. Sexual morality can lead to discrimination. Our sense of obligation can have perverse consequences. Bloom argues that morality, while 99% good, also fuels some of our worst excesses, and sometimes we would be better off "dialing down" our moral sense.

He contends that understanding evil requires recognizing that it's not always due to a lack of control, selfishness, or dehumanization. Sometimes, bad behavior is driven by a distorted moral sense.

Examples of "bad morality" include:

• Vengeance: The punitive impulse, while necessary for deterring bad behavior, can become indiscriminate, excessive, and shortsighted. Studies show people often prioritize maximum punishment even when it offers no future benefit or causes greater harm. This leads to cycles of violence, where each side believes they are acting righteously, as seen in conflicts like the Middle East.
• Moral Typecasting: Psychologist Kurt Gray's concept of "moral typecasting" highlights our tendency to categorize people as either "villains" or "victims." This binary thinking ignores the complexity of real-world situations where individuals can be both or neither. This leads to "competitive victimhood," where groups scramble for victim status to justify their actions and avoid accountability. This bias hinders productive dialogue and compromise, as each side believes they are unequivocally right and their opponents are wrong.

Bloom acknowledges that it's difficult to apply this perspective to current events, such as the trial of Luigi Manion, who murdered a healthcare CEO. While Bloom finds the act repellent, he recognizes that some people view it as a righteous act against a cruel system, driven by a moral sense as intense as his own.

Bloom is not advocating for a world without retribution. He believes it's essential for maintaining cooperation and deterring exploitation. However, he argues against an appetite for revenge that runs out of control, becoming unmoored from its legitimate purpose of deterrence. Our punitive desires can become excessive, seeking suffering for its own sake. A legal system, ideally, takes revenge out of individual hands and administers it fairly, providing a rational framework for justice that checks our unbridled retributive impulses.