AI Chatbots vs. Human Peers – Do People Prefer Chatbots Over Humans In Recovery?

Principal Category: Scam Victim Recovery Psychology / Recoverology

Authors:
•  Tim McGuinness, Ph.D. – Anthropologist, Scientist, Polymath, Director of the Society of Citizens Against Relationship Scams Inc.

Based on a study by Ruo-Ning Li, Dunigan Folk, Abhay Singh, Lyle Ungar, Elizabeth Dunn 2026

Abstract

AI Chatbots vs. Human Peers. Research comparing human peer interaction with supportive AI chatbots found that only human connection produced meaningful reductions in loneliness over time, despite chatbots temporarily improving mood. The findings are especially significant for traumatized scam victims, whose recovery is shaped by betrayal trauma, shattered trust, shame, cognitive impairment, and social withdrawal. Human support provides reciprocity, emotional co-regulation, vulnerability, and shared meaning that artificial systems cannot genuinely reproduce. While AI may offer immediate emotional comfort, psychoeducation, journaling prompts, or crisis triage support, it does not rebuild the relational trust damaged by scams. Overreliance on chatbot interaction may unintentionally deepen isolation by replacing difficult but necessary human engagement. Effective scam victim recovery, therefore, depends on trauma-informed human support groups, peer connection, and authentic interpersonal relationships as the foundation of long-term healing.

The Unbridgeable Gap: Applying Human-AI Interaction Research to Traumatized Scam Victims in Support Groups

Studying Loneliness

The global epidemic of loneliness, declared a public health concern by the U.S. Surgeon General and the World Health Organization, has found a proposed technological solution in AI companions. A recent pre-registered study by Li et al. (2026) offers a timely and critical investigation into this proposed remedy, comparing the efficacy of a highly supportive chatbot (“Sam”) with that of a random human peer in alleviating loneliness over a two-week period. Their findings, that only human interaction yielded a significant reduction in loneliness, resonate deeply within the context of trauma support, particularly for scam victims. This analysis will review the study’s methodology and findings, apply them to the unique psychological landscape of traumatized scam victims, and argue that while chatbots may offer a momentary balm, they are fundamentally inadequate and potentially counterproductive in the critical early days and months of recovery when genuine human connection is not just beneficial but essential for healing.

Review of the Li et al. (2026) Study

The study’s design was robust. The researchers recruited 296 first-year university students, a population experiencing a significant life transition and heightened loneliness, and randomly assigned them to one of three conditions for 14 days: interacting with a custom-built, highly supportive chatbot; interacting with a randomly assigned human peer; or a control condition of brief daily journaling. The chatbot “Sam” was thoughtfully designed, powered by ChatGPT-4o mini and programmed with principles from relationship science to be actively empathetic and validating.

The results were unambiguous. While participants in both the AI and human conditions reported reduced negative mood compared to the control group, only those who interacted with a human peer showed a statistically significant decrease in loneliness over the two-week period. The AI group’s loneliness scores were statistically indistinguishable from those who simply journaled. The study also found that participants in the human condition were more than twice as likely to continue the interaction after the study concluded, and a third exchanged contact information, suggesting a more durable and meaningful connection was formed. The researchers posited that the chatbot’s inability to reciprocate vulnerability and its lack of genuine, lived experience were key limiting factors. While the AI could simulate empathy, it could not provide the mutual responsiveness and co-construction of shared meaning that are foundational to human bonding.

The Unique Psychological Profile of the Traumatized Scam Victim

To apply these findings to scam victim support, one must first appreciate the profound and specific nature of their trauma. A scam victim’s experience is not merely a financial loss; it is a complex psychological injury characterized by:

1. Profound Betrayal and Shattered Trust: Unlike many other traumas, scams often involve a period of manufactured intimacy and trust. The victim’s ability to trust their own judgment and trust others is fundamentally shattered. The core wound is not just loss, but betrayal.
2. Intense Shame and Self-Blame: Scam victims are uniquely susceptible to overwhelming shame. They often internalize the societal narrative that they were “greedy,” “stupid,” or “naive,” leading to a profound sense of personal failure. This shame is a major barrier to seeking help, as they fear the very judgment they are already directing at themselves.
3. Neurological Impairment: As discussed in previous literature on trauma recovery, the post-scam brain is often in a state of hyperactivation. The prefrontal cortex, responsible for rational thought and decision-making, is impaired, while the amygdala, the fear center, is in overdrive. This results in reactive, impulsive behavior and an inability to process information logically.
4. Social Isolation (Exulansis): Victims frequently experience “exulansis”, the tendency to give up talking about their experience because they feel others cannot or will not understand its depth. They fear being dismissed, minimized, or judged, leading to a retreat into silence and deepening their isolation.

Application to the Early Days of Recovery: Why AI Fails Where Humans Succeed

Applying the Li et al. study to this specific population reveals that its conclusions are not just relevant, but amplified. In the first few days and weeks after discovering a scam, a victim’s needs are acute and specific. The study’s findings suggest that a chatbot would be worse than useless in this critical window; it could be actively harmful.

The Imperative of Validating Shame:

A scam victim’s most immediate need is the validation of their shame. They need to hear, unequivocally and from another human being, “This was not your fault. You were deceived by a sophisticated criminal.” While a chatbot like “Sam” can be programmed to say these words, the delivery lacks the essential weight of human conviction. As Perry (2023) and Shteynberg et al. (2024) argue, AI-generated empathy can feel empty upon reflection. For a victim drowning in self-blame, a human peer’s shared gasp of understanding or a simple, “I can’t imagine how you feel, but I’m here,” carries a validating power that a programmed response cannot replicate. The human peer’s own potential vulnerability and shared humanity create a relational bridge that the AI, with its perfect, un-lived empathy, cannot cross. The study found that participants felt more “heard” by their human partner, a feeling that is the direct antidote to the shame and isolation of exulansis.

The Necessity of Reciprocity and Co-regulation:

The Li et al. study highlights that human relationships are built on reciprocal patterns of self-disclosure. In the early days of trauma, a victim is not just a receiver of support; they are a volatile system in desperate need of co-regulation. A human peer, even another struggling victim, provides this through their own nervous system. They can sit with the victim in silence, mirror their calm, or share their own small piece of vulnerability, which signals safety and mutual understanding. A chatbot, by its very nature, creates a one-way street. It is a black hole of need. The victim can pour out their terror, rage, and grief, but the bot cannot reciprocate with a genuine story of its own fear or a shared moment of human fallibility. This one-directional flow can, over time, reinforce the victim’s sense of isolation and otherness. They are “cared for” by a machine, but they are not in connection with another being. This fails to rebuild the social and trust-related neural pathways that were damaged by the scam.

The Danger of an Illusion of Connection:

The Li et al. study suggests that while AI provides momentary emotional relief (a reduction in negative mood), this does not translate to long-term reductions in loneliness. For a traumatized scam victim, this is a critical distinction. A chatbot might provide a temporary distraction from the pain, much like a comfort movie or a video game. However, by providing an illusion of connection without the substance, it may inadvertently prevent the victim from engaging in the difficult but necessary work of forming real human bonds. It becomes a crutch that allows them to avoid the perceived risk of human judgment. The study’s finding that AI interaction did not reduce loneliness, while human interaction did, indicates that this crutch ultimately does not support the structure of long-term recovery. It may, in fact, exacerbate the core problem by deepening the victim’s withdrawal from the very human connections that could heal them.

The Role of AI in a Comprehensive Support Ecosystem: A Cautious Reassessment

This analysis should not be interpreted as a blanket condemnation of AI in the support ecosystem. Rather, it is an argument for its proper placement. Based on the Li et al. study and the psychology of trauma, chatbots are likely ill-suited as a primary support mechanism in the immediate aftermath of a scam. However, they may have utility in other, more ancillary roles:

• A Triage Tool: An AI could serve as a first point of contact on a support website, providing immediate, 24/7 information and directing the user to human resources. It can answer factual questions (“What do I do now?”) without the emotional weight of a human conversation.
• A Supplemental Journaling Aid: For a victim who is not yet ready to speak to anyone, a structured AI interaction might be a step up from a blank page. It could prompt them with gentle, non-judgmental questions to help them begin to articulate their experience, serving as a bridge to eventually speaking with a human. However, this must be framed clearly as a preparatory tool, not a replacement for connection.
• A Resource for Psychoeducation: Once a victim is further along in their recovery and their prefrontal cortex is back online, an AI can be an excellent source for information about trauma, common scams, and recovery strategies, free from the emotional labor required of a human support group member.

Conclusion: The Irreplaceable Human Element

The study by Li et al. (2026) provides crucial empirical evidence for a truth that trauma survivors and support professionals have long known intuitively: genuine connection heals. While technology can simulate empathy, it cannot replicate the shared vulnerability, mutual responsiveness, and embodied validation that are the hallmarks of a human relationship. For a traumatized scam victim in the first days and months of their recovery, these are not luxuries; they are necessities.

The fundamental question is whether chatbots are better or worse than humans in this critical period. Based on the study’s findings and the specific psychological needs of scam victims, the answer is clear: they are significantly worse. A chatbot offers a hollow echo of connection, a temporary palliative that may ultimately deepen the wound of loneliness it purports to heal. A human peer, even a random one, offers a mirror, a shared space of vulnerability, and a tangible reminder that the victim is not alone in their humanity. In the war against the isolation wrought by trauma, the imperfect, unpredictable, and profoundly human connection is, and will remain, the most effective weapon we have. Support groups must therefore continue to center human interaction as their primary and most valuable tool, using technology only to facilitate, never to replace, the irreplaceable power of one human being reaching out to another.

Study Document

Glossary

• AI Companion — An AI companion is a chatbot or artificial system designed to simulate social connection, emotional support, or companionship. In the article, this concept is examined through the comparison between a supportive chatbot and a human peer. For scam victims, an AI companion may provide temporary comfort, but it does not provide genuine human reciprocity or shared experience.
• AI-Generated Empathy — AI-generated empathy refers to supportive language produced by a chatbot that appears caring, validating, and emotionally responsive. The article explains that this form of empathy may feel comforting in the moment but lacks lived emotional reality. For scam victims, this limitation matters because recovery often requires authentic human witnessing, not only well-phrased reassurance.
• AI Support Ecosystem — An AI support ecosystem refers to the limited and carefully defined use of artificial intelligence within a broader recovery structure. The article presents AI as potentially useful for triage, journaling prompts, and psychoeducation. It warns that AI should support access to human care rather than replace human connection.
• Ancillary Role — An ancillary role is a secondary support function that assists but does not replace the primary healing relationship. In the article, chatbots are described as potentially useful in this limited role. For scam victims, ancillary tools may help organize information, but human support remains central to recovery.
• Chatbot Sam — Chatbot Sam was the custom-built AI companion used in the Li et al. study. It was designed to be supportive, validating, and rooted in relationship science principles. Its failure to reduce loneliness as effectively as human peer contact shows that supportive responses alone are not the same as genuine relationship.
• Chatbot Substitution Risk — Chatbot substitution risk refers to the danger that artificial interaction may replace needed human contact. The article warns that a chatbot can become a crutch that helps victims avoid the risk of human disclosure. This can deepen isolation when recovery requires safe, trauma-informed connection with real people.
• Co-Construction of Shared Meaning — Co-construction of shared meaning refers to the way two people build understanding together through mutual exchange. The study identifies this as a key feature of human relationship that chatbots cannot genuinely reproduce. For scam victims, shared meaning helps transform isolation into recognition, validation, and recovery participation.
• Control Condition — A control condition is the comparison group used in a study to help determine whether an intervention produces a meaningful effect. In the Li et al. study, the control group wrote brief daily journal entries. The chatbot group did not reduce loneliness more than this journaling condition, which limits claims about chatbot effectiveness.
• Daily Text Interaction — Daily text interaction refers to the repeated messaging activity used in the study over a two-week period. Participants either texted a human peer, interacted with a chatbot, or completed brief journaling. The results showed that daily human texting produced stronger loneliness reduction than daily chatbot interaction.
• Durable Connection — Durable connection refers to a relationship or interaction pattern that continues beyond an immediate moment of comfort. In the study, human participants were more likely to continue contact after the study ended. For scam victims, a durable connection matters because recovery requires ongoing trust repair and social reintegration.
• Early Recovery Window — The early recovery window refers to the first days, weeks, and months after a scam is discovered. The article describes this period as especially vulnerable due to shame, shock, isolation, and impaired decision-making. During this time, human support is more appropriate than chatbot companionship as the primary recovery response.
• Embodied Validation — Embodied validation refers to reassurance that comes from a real person with lived emotion, presence, and human responsiveness. The article contrasts this with simulated empathy produced by a chatbot. For scam victims, embodied validation can help counter shame because it comes from another human being who chooses to witness the pain.
• Emotional Balm — Emotional balm refers to temporary relief from distress, sadness, or loneliness. The article suggests that chatbots may provide this kind of short-term comfort. However, emotional balm should not be mistaken for the deeper relational repair needed after betrayal trauma caused by scams.
• Emotional Co-Regulation — Emotional co-regulation refers to the calming influence that can occur when one person’s nervous system helps another person stabilize. In support groups, a calm and compassionate human presence can help reduce fear and shame. Chatbots cannot provide true nervous system co-regulation because they do not possess embodied human presence.
• Emotional Weight — Emotional weight refers to the depth and seriousness carried by a human expression of care. A message from another person can feel meaningful because that person has chosen to respond. The article explains that chatbot responses may lack this weight because they are generated rather than personally lived.
• Exulansis — Exulansis refers to the tendency to stop talking about an experience because others do not seem able or willing to understand it. The article applies this concept to scam victims who withdraw after judgment, minimization, or misunderstanding. Human support groups can reduce exulansis by creating shared recognition and safe disclosure.
• Feeling Heard — Feeling heard refers to the perception that another person has paid attention, understood, and responded meaningfully. The study measured this as a possible mediator between interaction type and social benefit. For scam victims, feeling heard is central because shame and isolation often grow when others dismiss or misunderstand the crime.
• First-Year University Transition — First-year university transition refers to the life change experienced by the study participants. The researchers selected this group because it often involves increased loneliness and the need to build new social ties. The article uses this population carefully while applying the findings to scam victims with stronger trauma-related needs.
• Genuine Human Connection — Genuine human connection refers to real interpersonal contact involving choice, presence, vulnerability, and mutual response. The article argues that this connection cannot be fully reproduced by artificial systems. For scam victims, a genuine connection helps repair the relational injury created by deception and manufactured intimacy.
• Human Peer Interaction — Human peer interaction refers to a supportive exchange with another person who is not acting as an artificial companion. In the study, human peer interaction reduced loneliness more effectively than chatbot interaction. In scam recovery, peer contact can help victims feel less alone and more understood.
• Human Vulnerability — Human vulnerability refers to the ability of a person to share limitation, uncertainty, emotion, or lived experience. The article identifies this as a feature missing from chatbot interaction. Scam victims benefit from human vulnerability because it creates mutuality rather than one-directional support.
• Illusion of Connection — Illusion of connection refers to the feeling of being emotionally supported without the presence of a genuine human relationship. The article warns that chatbots may create this temporary impression. For scam victims, this illusion can be risky because the original trauma often involved false intimacy and a deceptive connection.
• Interpersonal Closeness — Interpersonal closeness refers to the felt sense of emotional nearness or relationship connection with another person. The study examined closeness as one possible factor in loneliness outcomes. For scam victims, closeness must be rebuilt carefully because betrayal trauma can damage the ability to trust a connection.
• Loneliness Reduction — Loneliness reduction refers to a measurable decrease in the feeling of social disconnection. In the study, only the human peer condition produced a significant reduction in loneliness over time. This finding supports the importance of real human contact in recovery environments where isolation is a major concern.
• Manufactured Intimacy — Manufactured intimacy refers to false emotional closeness created through manipulation, deception, or scripted affection. Romance scams often rely on this process to build trust and dependence. The article’s concern about chatbots is heightened because scam victims have already been harmed through artificial or deceptive intimacy.
• Momentary Negative Mood Relief — Momentary negative mood relief refers to a short-term reduction in sadness, distress, or emotional discomfort. The study found that chatbot interaction could reduce negative mood compared with journaling. The article explains that this kind of relief does not equal durable loneliness reduction or relational healing.
• Mutual Responsiveness — Mutual responsiveness refers to reciprocal attention, care, disclosure, and emotional exchange between people. Relationship science identifies this as a foundation of intimacy and connection. Chatbots may respond supportively, but they cannot participate in true mutual responsiveness because they do not have lived vulnerability.
• One-Directional Support — One-directional support refers to a support pattern in which one side receives expression while the other side does not genuinely share or need care. The article describes chatbot interaction as tending toward this one-way structure. For scam victims, one-directional support may soothe temporarily, but does not rebuild relational confidence.
• Peer Support Groups — Peer support groups are human-centered recovery settings where individuals with related experiences share, listen, and learn from one another. The article argues that these groups are essential for traumatized scam victims. They provide recognition, belonging, mutual understanding, and a structured human connection that chatbots cannot replace.
• Pre-Registered Study — A pre-registered study is research in which hypotheses, methods, and analysis plans are recorded before data are examined. The Li et al. study used this design, which strengthens confidence in its findings. The article relies on this study because it directly compared human peer contact, chatbot interaction, and journaling.
• Primary Support Mechanism — A primary support mechanism is the main source of recovery assistance used during an acute or vulnerable period. The article argues that chatbots should not serve this role for newly harmed scam victims. Human contact should remain primary because early recovery requires validation, safety, and relational repair.
• Profound Betrayal — Profound betrayal refers to the deep violation of trust created by manipulation, deception, and emotional exploitation. The article identifies betrayal as a core feature of scam victim trauma. This betrayal damages both trust in others and trust in personal judgment, making human support especially important.
• Psychoeducation Resource — A psychoeducation resource provides information about trauma, recovery, scams, and coping strategies. The article recognizes AI as potentially useful for this purpose when victims are stable enough to process information. Psychoeducation should support recovery, but it should not substitute for human care during early distress.
• Random Human Peer — A random human peer is an ordinary person assigned for interaction without specialized therapeutic training. In the study, random human peers were more effective than the supportive chatbot in reducing loneliness. This finding suggests that even imperfect human presence can carry benefits that polished artificial support does not provide.
• Reciprocal Self-Disclosure — Reciprocal self-disclosure refers to the mutual sharing of personal experiences, feelings, or vulnerabilities. Human relationships often deepen through this exchange. Chatbots cannot genuinely disclose lived experience, which limits their ability to support deeper relational healing for scam victims.
• Relational Bridge — A relational bridge is the connection that allows one person to feel met, recognized, and accompanied by another. The article uses this idea to explain why human validation has greater recovery value. For scam victims, the relational bridge helps counter isolation, shame, and the belief that no one can understand.
• Relational Security — Relational security refers to the sense of safety that develops when connection is trustworthy, consistent, and mutual. Scam victims often lose this sense after being manipulated through false closeness. Human support groups can help rebuild relational security through real presence, respectful boundaries, and repeated safe interaction.
• Robust Study Design — A robust study design refers to research methods that strengthen confidence in the findings. The Li et al. study included random assignment, multiple conditions, pre- and post-measures, and a control group. This makes the findings especially useful when considering support models for vulnerable groups.
• Shared Humanity — Shared humanity refers to the recognition that another person is present, imperfect, vulnerable, and emotionally real. The article presents shared humanity as something a chatbot cannot genuinely provide. For scam victims, shared humanity helps reduce shame by reminding them that they are not alone or uniquely defective.
• Shattered Trust — Shattered trust refers to the collapse of confidence in others and in one’s own judgment after deception. The article identifies this as one of the central injuries in scam victim trauma. Recovery requires careful rebuilding of trust through safe human contact, not merely informational reassurance.
• Social Isolation — Social isolation refers to withdrawal from human contact due to shame, fear, mistrust, or the belief that others will not understand. Scam victims often experience this after disclosure or discovery of the crime. Human-centered support helps interrupt isolation by creating safe spaces for connection and understanding.
• Social Support Measurement — Social support measurement refers to the study’s effort to assess participants’ perceived support and relationship satisfaction. The study examined this alongside loneliness, mood, perceived isolation, and closeness. These measures help distinguish short-term emotional comfort from broader social and relational benefit.
• Supplemental Journaling Aid — A supplemental journaling aid is a tool that helps a person begin writing, reflecting, or organizing emotional material. The article suggests that AI may serve this limited role for victims not ready to speak with others. It must be framed as preparation for human support rather than a replacement for it.
• Support Group Centering — Support group centering refers to keeping human interaction at the core of recovery programming. The article concludes that support groups should use technology only to facilitate human connection. This approach protects scam victims from relying on artificial companionship when an authentic relationship is needed.
• Technology-Facilitated Support — Technology-facilitated support refers to the use of digital tools to improve access to recovery resources. The article allows for technology when it directs victims toward information, journaling, or human help. It rejects technology as a substitute for the human presence required in early trauma recovery.
• Temporary Palliative — A temporary palliative is a short-term comfort that reduces distress without resolving the underlying wound. The article describes chatbot interaction in this way when it provides emotional relief but does not reduce loneliness. Scam victims may benefit from temporary relief, but recovery requires deeper human reconnection.
• Triage Tool — A triage tool is a first-contact resource that helps sort urgent needs and guide a victim toward appropriate help. The article suggests that AI may answer basic questions and direct victims to human resources. This use is safer when the tool clearly leads toward real support rather than prolonged artificial companionship.
• Trust-Related Neural Pathways — Trust-related neural pathways refer to brain and nervous system patterns involved in safety, connection, and relational expectation. The article argues that scam trauma damages these pathways through betrayal and manufactured intimacy. Human connection is needed to help rebuild these patterns through safe relational experience.
• UCLA Loneliness Scale — The UCLA Loneliness Scale is a standardized measure used in the Li et al. study to assess loneliness. Participants completed it before and after the two-week intervention period. The results showed that human peer interaction reduced loneliness while chatbot interaction did not significantly differ from journaling.
• Validation of Shame — Validation of shame refers to acknowledging the victim’s painful self-blame while correcting the false belief that the crime was their fault. The article identifies this as an immediate need for scam victims. Human validation carries special power because it comes from another person’s presence, conviction, and willingness to witness.
• Victim Self-Blame — Victim self-blame refers to the belief that the scam occurred because of personal failure, weakness, greed, or stupidity. The article identifies this as a common and damaging reaction among scam victims. Trauma-informed support must correct this belief by naming manipulation, criminal sophistication, and the reality of coercive deception.
• Withdrawal From Human Connection — Withdrawal from human connection refers to retreating from people because of shame, fear, or expected judgment. The article warns that chatbot dependence may reinforce this withdrawal by providing private comfort without relational risk. Recovery requires careful return to safe human contact, not deeper isolation behind artificial reassurance.