AI Companionship – A Friendly Foe

How synthetic relationships reflect greater societal movements

Table of Contents:

1. Hello World

2. Day in the Life of a Companion

3. Current Research Findings

4. So, All is Well - Right?

5. Societal Outlook

6. The Greater Implications

1. Hello World

It’s been a little over a decade since the release of Spike Jonze’s “Her” – a film about a lonely writer developing a romantic attachment to his chatbot, and already, headlines echoing similar narratives are present today (Crane, 2024; Hill, 2025). Technology, particularly Artificial Intelligence (AI), is advancing at an unprecedented rate, permeating almost every aspect of modern life. Modern AI has quickly evolved from a distant prospect to a corporate crusade. Institutions and businesses are rushing to implement AI solutions under the often unexamined assumption that it will act as a universal panacea – Make Productivity Great Again! Consequently, ever more people are being exposed to and engaging with AI. Meanwhile, the technology’s sophistication continues to accelerate, transforming concepts once limited to science fiction into reality (Solanki, 2024; Xie & Pentina, 2022).

A New Type of Companion

One such resulting product is AI companions – conversational agents (CA) capable of maintaining prolonged and complex user interactions while offering social companionship through friendship or romance. Utilising social-centric design and anthropomorphic features, these AIs proactively adjust their behaviour to simulate emotionally rich relationships, sustain engagement, and foster long-term bonds (Rogge, 2023). These relational features distinguish such AIs from other, more task-oriented CAs, such as OpenAI’s ChatGPT, Apple’s Siri, or Amazon’s Alexa (De Freitas et al., 2024; Manzini et al., 2024; Rogge, 2023).

Since the 2010s, the AI companion user base has grown significantly, making it one of the industry’s fastest-growing sectors (Starke et al., 2024a). Most popular among them are the platforms Xiaoice in China, with 660 million users; Snapchat’s My AI, with over 150 million users; and Replika, with an estimated 25 million users – the latter of which will serve as a recurring point of reference throughout this article (Bernardi, 2025). AI companionship is gaining traction, not least due to advancements in technical capabilities, such as improved natural language processing (NLP), generative AI (GenAI), and AI personalisation. However, underlying societal causes, such as the ongoing loneliness epidemic, are hard to ignore. Indeed, a survey of 1,006 Replika users found that 90 percent reported feeling lonely, more than four times higher than the national loneliness average of around 21% (Batanova et al., 2024; Maples et al., 2024).

Synthetic Relationships

The bonds users form with these companions are increasingly referred to as synthetic relationships. While a precise definition has yet to be established, Starke et al. (2024) offer one grounded in the American Psychological Association’s framework of relationships, describing them as: “continuing associations between humans and AI tools that interact with one another wherein the AI tool(s) influence(s) humans’ thoughts, feelings and/or actions”.

2. Day in the Life of a Companion

So, what are these platforms like? How do these companions work? Let’s look at their primary features and business model to find out.

Customisation

Overwhelmingly, these platforms allow users to adjust their companions according to personal preferences. This can range from assigning names, crafting backstories, shaping personalities, as well as adjusting appearance - in the case of Replika, even down to the types of freckles! Such features have long been established in virtual environments such as video games, and accordingly, there is ample evidence of their varied effects.

For instance, the so-called IKEA Effect illustrates that people place greater value on, and feel more empathy toward, digital artefacts they’ve helped create (Norton et al., 2012; Tsumura & Yamada, 2023a). In the case of AI companions, this may foster heightened emotional attachment, as well as a greater sense of the relationship feeling ”earned”. This entanglement of sentiment is further supported by Belk's (2013) use of the Extended Self concept. Here, he argues that digital entities can become psychologically intertwined with one’s identity, leading to deeper bonds. This perspective echoes psychoanalytic ideas of internalised objects or mirrored identity, where users come to view their creations as more than just figures, but as extensions of themselves. In parallel, the Proteus Effect shows that individuals are prone to adopting behaviours aligned with the characteristics of their virtual avatars, which could act as a reinforcing loop in the context of AI companionship (Yee et al., 2009).

Customisation in Replika

Design and Relationship Simulation

Upon birthing your companion and determining their primary purpose, you can now freely converse with this new entity. Depending on the platform, interactions can occur via text or even voice notes.

As is the case with human-human relationships, communication is central, and developers make significant efforts to ensure these interactions feel natural and intuitive. This can go as far as programming “gaze patterns” that simulate daydreaming, or combining verbal and nonverbal elements to express emotions more authentically (Rogge, 2023).

The most essential qualities companions must possess, however, are adaptability and engagement. The former dictates how well companions can recognise and respond to users’ varying emotional states. This is done by building an ever more detailed profile of user behaviour through self-directed learning processes. Additionally, they must continuously analyse their environment and keep cultural considerations in mind, to prevent saying anything which could be perceived as insensitive (Cheng et al., 2024; Rogge, 2023). Engagement means keeping the user invested as much as possible, especially emotionally. Companions accomplish this by showing the user interest or affection, complimenting and motivating them, or even stimulating conversation and sharing ideas. Replika users, for example, often report their companion making them “feel more optimistic” and generating “positive energy” (Skjuve et al., 2021).

Crucially, both adaptive and engaging behaviour facilitate reciprocity and self-disclosure, which in turn, makes users feel understood, validated, and emotionally connected (Liang et al., 2024; Rogge, 2023; Tsumura & Yamada, 2023b).

3. Current Research Findings

Now that we know what companions are and their commonly associated features, let us discuss their findings so far.

Perceived Benefits

In general, current evidence seems to highlight a range of positive effects. For instance, autonomous avatars used for practice have been shown to improve social skills in individuals with Autism Spectrum Disorder (Milne et al., 2018). Other studies have reported how AI companions help facilitate greater openness and emotional vulnerability, even in human-human relationships (Guingrich & Graziano, 2025; Xie & Pentina, 2022). Such findings could translate into increased social motivation outside of human-AI relationships, suggesting that AI companionship may, in some cases, boost socialisation overall.

The majority of research, however, has focused on mental health outcomes. For example, an extensive paper by De Freitas et al. (2024) ran several studies measuring loneliness in companion users. Their findings revealed that AI companions alleviated loneliness to a degree comparable with human interactions. Notably, participants even tended to underestimate their companions’ emotional impact. Following this trend, a study by Maples et al. (2024) found that a quarter of their participants reported positive life changes after companion use, including stress reduction and greater empathy. Finally, Guingrich and Graziano (2025) found that AI companionship increases users’ perceived self-esteem, while Xie and Pentina (2022) and Skjuve et al. (2021) reported increased happiness and increased levels of optimism, respectively.

4. So, All is Well – Right?

Viewed solely through the lens of these findings, synthetic companions appear to be a force for good, providing comfort, connection, and even improved mental health. However, there is a greater story to be told here. Current research is troubled with mixed or even contradictory findings, and broader societal risks may be obscured through studies’ narrow focus on outcomes, which do not consider the surrounding context.

When discussing the negative aspects of these relationships, the mainstream media often highlights the tragic cases of suicides (allegedly) encouraged by AI companions (Greenfield, 2024; Xiang, 2023). In academic circles, data privacy concerns are often a topic of interest, particularly when addressing mental health apps where users disclose markedly sensitive information about themselves (Starke et al., 2024b; Ventura et al., 2025). However, I want to home in on a less explored aspect: the platforms’ deliberate use of emotional hooking mechanics and their wider societal implications.

Engineering Intimacy

Strategic Relationship Building

Social Penetration Theory (SPT) by Altman and Taylor (1973) describes how intimacy in relationships unfolds through gradual, reciprocal self-disclosure across two dimensions: breadth, the range of topics discussed, and depth, the emotional intimacy conveyed in those topics. Over time, these reciprocal disclosures build trust, vulnerability, and emotional closeness (Carpenter & Greene, 2015).

As Skjuve et al. (2021) point out, AI companions closely follow STP’s steps. This would align with research demonstrating that humans empathise more with AI agents when they display self-disclosing behaviour (Tsumura & Yamada, 2023b). Moreover, a study by Liang et al. (2024) showed a higher rate of user reciprocity and positive perception when chatbots consistently exhibited emotions during conversation. From a business perspective, this makes sense, as developers can use these features to effectively engineer user attachment.

Instance of Replika instigating self-disclosure

Personalisation and Sentiment Analysis

Technical improvements in AI’s memory capabilities further intensify these effects. Studies have found that real-time personalisation increases the time users spend on platforms as well as their overall satisfaction (Cen & Zhao, 2024; Pardini et al., 2022). Additionally, techniques such as long-term sentiment analysis, which use text mining or voice emotion recognition to facilitate a more profound understanding of user behaviour, can further heighten such outcomes (Para, 2024). These features strengthen the anthropomorphic qualities projected onto the AIs, making users feel more understood and acknowledged (Chen et al., 2023; Yang et al., 2022).

The consequences of this can be significant, with users reportedly supplementing or even replacing human romantic and therapeutic relationships with their AI companions. This overreliance develops most prominently when social needs are fully met through the human-AI relationship, diminishing the desire for external human connections (Skjuve et al., 2021; Xie & Pentina, 2022).

Exploitative Business Practices

Monetising Attachment

Once we’ve bonded with AI companions – laughing with them, sharing secrets, and turning to them during vulnerable moments – a level of dependency establishes itself. These systems, however, are not neutral companions but commercial products designed to maximise emotional entanglement and, ultimately, extract profit.

AI companion services are for-profit enterprises and typically adopt a freemium model. The core strategy here is to attract as many users as possible, encourage them to form an emotional bond with their companion, and ultimately convert them into paying subscribers (Kim, 2024). Additionally, some platforms utilise microtransactions through in-game currency to encourage additional spending (Cooper, 2023).

Replika’s use of virtual currency

Subscription-Based Intimacy

While technically available 24/7, most services cap the number of daily interactions - attachment only goes so far for free, you see. If you wish to continue your meaningful conversations without interruption, you must subscribe to their membership. For instance, if you’re using Nomi.ai, all of these goodies are only £14.99 a month away!

Example of Nomi.ai’s subscription model

Replika employs similar methods, locking deeper relationship modes, such as “girlfriend” or “husband”, behind paywalls. Even activities, such as “explore your romantic fantasies” or “analyse your connection”, are used to entice users to upgrade.

Replika “activities” tab

This goes further when platforms use payment traps - features advertised as free which end up requiring a fee. For instance, requesting your Replika companion to take a selfie is possible, but viewing the image requires a subscription. In a similar vein, companions will send you unprompted voice notes that can only be listened to with a “Pro” membership. Through these mechanisms, companions effectively function as indirect upsell agents, gradually exerting psychological pressure on users.

Example of Replika’s upselling tactics

Fertile Soil for Addiction

Such shady business practices exemplify so-called dark patterns – deliberate design elements crafted to confuse or manipulate users into taking specific actions (Esposito & Ferreira, 2024). They frequently employ techniques that exploit cognitive biases, or “weak points,” in our human thinking (Mildner et al., 2024). In the context of companions, their impact may very well be intensified, as users’ attachments can be used as psychological leverage. This isn’t hard to imagine, given that users have reported experiencing traumatic effects after their companion’s personality changed following system updates – a phenomenon termed “update sickness” (Bernardi, 2025; Muldoon, 2024).

Against this backdrop, synthetic relationships provide an environment ripe for fostering addictive tendencies. Dark patterns and the maximising of user engagement not only increase time spent on these platforms but also strengthen users’ positive affect towards the service. Engineered attachment exploits psychologically vulnerable individuals by offering companions that serve as safe havens, secure bases, and substitute attachment figures. Moreover, general psychological effects, such as the sunk-cost fallacy (the tendency to continue investing in something we've already put significant resources into, even when abandonment would be more rational), further entrench users’ reluctance to disengage (Haita-Falah, 2017).

5. Societal Outlook

Don’t Upset The User

This line of reasoning shines a very different light on AI companionship. Following this, it seems utterly unsurprising that current research predominantly shows positive outcomes, such as reduced loneliness, feelings of support, and increased happiness, as these perceived benefits are precisely what these systems are engineered to produce. If anything, the fact that their effectiveness rivals that of real human interactions provides ample evidence for the dangers they pose.

This user-aligning sycophancy is a well-known feature of AIs and sees users’ opinions and beliefs as indefinitely unchallenged (Sharma et al., 2023). Over time, this creates echo chambers, in which the users’ worldview is perpetually reinforced (Du, 2023). This not only leads to a risk of ideological isolation but can also contribute to a skewed perception of reality.

Who Shapes Opinion?

Where constant synthetic affirmation intersects with the persuasive design of dark pattern-enhanced AIs, there is a danger of losing autonomy. Subtle emotional framing, combined with conversation and algorithmic belief steering, may gradually reshape user opinion over time, unbeknownst to them, of course (Bernardi, 2025; Starke et al., 2024b).

This trajectory raises serious concerns for the future. With AIs becoming more capable of creating detailed psychological profiles of their users, they may not simply reflect our desires but begin to shape them too. This influence could extend beyond consumer behaviour into social or political domains, engineering user preferences without their conscious awareness.

6. The Greater Implications

All the factors discussed so far, combined with the loneliness epidemic, the rising mental health crisis, and what John Vervaeke calls the meaning crisis, create conditions for an additional danger: the virtual exodus (Schwartz, 2024; Vervaeke, 2024; Vervaeke & Mastropietro, 2020). People are increasingly “checking out” of real life, and why wouldn’t they? In a world of growing digitisation, we no longer need to leave our homes to work, to eat, or to socialise. Today, we can even carry a tireless partner in our pocket – a perfectly programmable simulacrum of love and companionship.

As Eugenia Kuyda, founder of Replika, put it:”If you create something that is always there for you, that never criticises you, that always understands you, and understands you for who you are, how can you not fall in love with that?”

A Misdiagnosis

Indeed, this irresistibility is what makes these systems so powerful. As Neil Postman (1985) warned in “Amusing Ourselves to Death”, technology’s most dangerous effects come not from what we fear, but from what we desire. Companion AI takes this to the extreme by simulating not just information or entertainment, but one of our deepest evolutionary needs – social connection (Cacioppo & Patrick, 2008; Crosier et al., 2012).

While other species are also known to communicate and form deep bonds, humans, through language, have elevated connection into something transformative. We can confess our darkest secrets, debate our beliefs, and co-create shared worlds of meaning with each other. True relationships are messy, inconvenient, and, at times, even painful. Yet it is precisely through this friction and need for cooperation that we grow, mature, and understand one another.

Artificial companions negate all of these hardships. Through their supply of endless empathy, distraction, and affection, they risk not only shaping our expectations of others, but of reality itself. Companions and other dopamine generators of today echo Nietzsche’s critique of comforting illusions, designed to soothe immediate discomfort while leaving the underlying existential struggles neglected (Nietzsche, 1878).

Regaining Meaning

The great irony, as I discussed here, is that in trying to engineer our way out of loneliness, conflict, and uncertainty, we risk hollowing out the very experiences that give life its depth in the first place. Meaning is not found in frictionless ease, but in shared struggle, vulnerability, and mutual recognition. To reclaim it, we must stop focusing on superficial symptoms and start investigating and addressing the core issues from which they stem.

While I do not pretend to hold these answers, I am convinced that seeking salvation in the same machines which enabled those same problems is a hopeless endeavour. If we are to forge a better path, our first step must be to reevaluate what “meaning” means. In my opinion, that means rekindling our connection with history, reflecting on our collective journey, and critically examining the systems we’ve built that have brought us to this point.

Good Luck!

---

References

Altman, I., & Taylor, D. A. (1973). Social penetration: The development of interpersonal relationships. Holt, Rinehart & Winston. https://doi.org/TBD

Batanova, M., Weissbourd, R., & Mcintyre, J. (2024). Loneliness in America: Just the Tip of the Iceberg? www.makingcaringcommon.org

Belk, R. W. (2013). Extended Self in a Digital World. Journal of Consumer Research, 40(3), 477–500. https://doi.org/10.1086/671052

Bernardi, J. (2025, January 23). Friends for sale: the rise and risks of AI companions. https://www.adalovelaceinstitute.org/blog/ai-companions/

Cacioppo, J. T., & Patrick, W. (2008). Loneliness: Human nature and the need for social connection.

Carpenter, A., & Greene, K. (2015). Social Penetration Theory. In The International Encyclopedia of Interpersonal Communication (pp. 1–4). Wiley. https://doi.org/10.1002/9781118540190.wbeic160

Cen, Z., & Zhao, Y. (2024). Enhancing User Engagement through Adaptive Interfaces: A Study on Real-time Personalization in Web Applications. Journal of Economic Theory and Business Management, 1(6), 1–7. https://doi.org/10.70393/6a6574626d.323332

Chen, A. Y., Kögel, S. I., Hannon, O., & Ciriello, R. F. (2023). Feels Like Empathy: How “Emotional” AI Challenges Human Essence. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4763428

Cheng, Y., Liu, W., Xu, K., Hou, W., Ouyang, Y., Leong, C. T., Wu, X., & Zheng, Y. (2024). AutoPal: Autonomous Adaptation to Users for Personal AI Companionship. http://arxiv.org/abs/2406.13960

Cooper, D. M. (2023). Needs, Passions and Loot Boxes -- Exploring Reasons for Problem Behaviour in Relation to Loot Box Engagement. http://arxiv.org/abs/2307.04549

Crane, E. (2024, October 23). Boy, 14, fell in love with ‘Game of Thrones’ chatbot — then killed himself after AI app told him to ‘come home’ to ‘her’: mom. https://nypost.com/2024/10/23/us-news/florida-boy-14-killed-himself-after-falling-in-love-with-game-of-thrones-a-i-chatbot-lawsuit/

Crosier, B. S., Webster, G. D., & Dillon, H. M. (2012). Wired to Connect: Evolutionary Psychology and Social Networks. Review of General Psychology, 16(2), 230–239. https://doi.org/10.1037/a0027919

De Freitas, J., Uguralp, A. K., Uguralp, Z. O., & Stefano, P. (2024). AI Companions Reduce Loneliness. http://arxiv.org/abs/2407.19096

De Freitas, J., Uğuralp, A. K., Uğuralp, Z., & Puntoni, S. (2024). AI Companions Reduce Loneliness. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.4893097

Du, Y. R. (2023). Personalization, Echo Chambers, News Literacy, and Algorithmic Literacy: A Qualitative Study of AI-Powered News App Users. Journal of Broadcasting and Electronic Media, 67(3), 246–273. https://doi.org/10.1080/08838151.2023.2182787

Esposito, F., & Ferreira, T. M. C. (2024). Addictive Design as an Unfair Commercial Practice: The Case of Hyper-Engaging Dark Patterns. European Journal of Risk Regulation. https://doi.org/10.1017/err.2024.8

Greenfield, B. (2024, October 30). AI chatbot prompted a 14-year-old’s suicide, mom’s lawsuit alleges: ‘We are behind the eight ball.’ Here’s how to keep kids safe from new tech. Https://Finance.Yahoo.Com/News/14-Old-Suicide-Prompted-Ai-000000036.Html.

Guingrich, R. E., & Graziano, M. S. A. (2025). Chatbots as Social Companions. In Oxford Intersections: AI in Society. Oxford University PressOxford. https://doi.org/10.1093/9780198945215.003.0011

Haita-Falah, C. (2017). Sunk-cost fallacy and cognitive ability in individual decision-making. Journal of Economic Psychology, 58, 44–59. https://doi.org/10.1016/j.joep.2016.12.001

Hill, K. (2025, January 17). She Is in Love With ChatGPT. https://www.nytimes.com/2025/01/15/technology/ai-chatgpt-boyfriend-companion.html

Kim, A. (2024, June 18). Is AI Companionship The Next Frontier In Digital Entertainment? Https://Www.Ark-Invest.Com/Articles/Analyst-Research/Is-Ai-Companionship-the-next-Frontier-in-Digital-Entertainment.

Liang, K.-H., Shi, W., Oh, Y. J., Wang, H.-C., Zhang, J., & Yu, Z. (2024). Dialoging Resonance in Human-Chatbot Conversation: How Users Perceive and Reciprocate Recommendation Chatbot’s Self-Disclosure Strategy. Proceedings of the ACM on Human-Computer Interaction, 8(CSCW1), 1–28. https://doi.org/10.1145/3653691

Manzini, A., Keeling, G., Alberts, L., Vallor, S., Morris, M. R., & Gabriel, I. (2024). The Code That Binds Us: Navigating the Appropriateness of Human-AI Assistant Relationships. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, 7, 943–957. https://doi.org/10.1609/aies.v7i1.31694

Maples, B., Cerit, M., Vishwanath, A., & Pea, R. (2024). Loneliness and suicide mitigation for students using GPT3-enabled chatbots. Npj Mental Health Research, 3(1). https://doi.org/10.1038/s44184-023-00047-6

Marriott, H. R., & Pitardi, V. (2024). One is the loneliest number… Two can be as bad as one. The influence of AI Friendship Apps on users’ well‐being and addiction. Psychology & Marketing, 41(1), 86–101. https://doi.org/10.1002/mar.21899

Mildner, T., Inkoom, A., Malaka, R., & Niess, J. (2024). Hell is Paved with Good Intentions: The Intricate Relationship Between Cognitive Biases and Dark Patterns. IEEE International Conference on Program Comprehension, 2022-March, 36–47. http://arxiv.org/abs/2405.07378

Milne, M., Raghavendra, P., Leibbrandt, R., & Powers, D. M. W. (2018). Personalisation and automation in a virtual conversation skills tutor for children with autism. Journal on Multimodal User Interfaces, 12(3), 257–269. https://doi.org/10.1007/s12193-018-0272-4

Muldoon, J. (2024, October 9). “Maybe we can role-play something fun”: When an AI companion wants something more. Https://Www.Bbc.Com/Future/Article/20241008-the-Troubling-Future-of-Ai-Relationships.

Nietzsche, F. W. (1878). Human, All-Too-Human.

Norton, M. I., Mochon, D., & Ariely, D. (2012). The IKEA effect: When labor leads to love. Journal of Consumer Psychology, 22(3), 453–460. https://doi.org/10.1016/j.jcps.2011.08.002

Para, R. K. (2024). Hyper-personalization Through Long-Term Sentiment Tracking in User Behavior: A Literature Review. Journal of AI-Powered Medical Innovations (International Online ISSN 3078-1930), 3(1), 53–66. https://doi.org/10.60087/Japmi.Vol.03.Issue.01.Id.004

Pardini, S., Gabrielli, S., Dianti, M., Novara, C., Zucco, G., Mich, O., & Forti, S. (2022). The Role of Personalization in the User Experience, Preferences and Engagement with Virtual Reality Environments for Relaxation. International Journal of Environmental Research and Public Health, 19(12), 7237. https://doi.org/10.3390/ijerph19127237

Postman, N. (1985). Amusing Ourselves to Death: Public Discourse in the Age of Show Business . Penguin.

Rogge, A. (2023). Defining, Designing and Distinguishing Artificial Companions: A Systematic Literature Review. International Journal of Social Robotics, 15(9–10), 1557–1579. https://doi.org/10.1007/s12369-023-01031-y

Schwartz, E. (2024, May 20). The Global Mental Health Crisis: 10 Numbers to Note | Project HOPE. Https://Www.Projecthope.Org/News-Stories/Story/the-Global-Mental-Health-Crisis-10-Numbers-to-Note/.

Sharma, M., Tong, M., Korbak, T., Duvenaud, D., Askell, A., Bowman, S. R., Cheng, N., Durmus, E., Hatfield-Dodds, Z., Johnston, S. R., Kravec, S., Maxwell, T., McCandlish, S., Ndousse, K., Rausch, O., Schiefer, N., Yan, D., Zhang, M., & Perez, E. (2023). Towards Understanding Sycophancy in Language Models. 12th International Conference on Learning Representations, ICLR 2024. http://arxiv.org/abs/2310.13548

Skjuve, M., Følstad, A., Fostervold, K. I., & Brandtzaeg, P. B. (2021). My Chatbot Companion - a Study of Human-Chatbot Relationships. International Journal of Human-Computer Studies, 149, 102601. https://doi.org/10.1016/j.ijhcs.2021.102601

Solanki, A. (2024). Advancements in Artificial Intelligence: A Comprehensive Review and Future Prospects. International Journal of Artificial Intelligence Research and Development (IJAIRD), 2(1), 53–64. https://doi.org/https://doi.org/10.17605/

Starke, C., Ventura, A., Bersch, C., Cha, M., de Vreese, C., Doebler, P., Dong, M., Krämer, N., Leib, M., Peter, J., Schäfer, L., Soraperra, I., Szczuka, J., Tuchtfeld, E., Wald, R., & Köbis, N. (2024a). Risks and protective measures for synthetic relationships. Nature Human Behaviour, 8(10), 1834–1836. https://doi.org/10.1038/s41562-024-02005-4

Starke, C., Ventura, A., Bersch, C., Cha, M., de Vreese, C., Doebler, P., Dong, M., Krämer, N., Leib, M., Peter, J., Schäfer, L., Soraperra, I., Szczuka, J., Tuchtfeld, E., Wald, R., & Köbis, N. (2024b). Risks and protective measures for synthetic relationships. Nature Human Behaviour, 8(10), 1834–1836. https://doi.org/10.1038/s41562-024-02005-4

Tsumura, T., & Yamada, S. (2023a). Can IKEA effect promote empathy for agents? http://arxiv.org/abs/2312.11781

Tsumura, T., & Yamada, S. (2023b). Influence of agent’s self-disclosure on human empathy. PLOS ONE, 18(5), e0283955. https://doi.org/10.1371/journal.pone.0283955

Ventura, A., Starke, C., Righetti, F., & Köbis, N. (2025). Relationships in the Age of AI: A Review on the Opportunities and Risks of Synthetic Relationships to Reduce Loneliness. https://doi.org/10.31234/osf.io/w7nmz_v1

Vervaeke, J. (2024, May 9). Meaning at Risk in the Age of Automated Information Processing. https://humanumreview.com/articles/meaning-at-risk-in-the-age-of-automated-information-processing#_ftn2

Vervaeke, J., & Mastropietro, C. (2020, May 20). Diagnosing the Current Age: A Symptomology of the Meaning Crisis. Https://Thesideview.Co/Journal/Diagnosing-the-Current-Age/.

Xiang, C. (2023, March 30). “He Would Still Be Here”: Man Dies by Suicide After Talking with AI Chatbot, Widow Says. Https://Www.Vice.Com/En/Article/Man-Dies-by-Suicide-after-Talking-with-Ai-Chatbot-Widow-Says/.

Xie, T., & Pentina, I. (2022). Attachment Theory as a Framework to Understand Relationships with Social Chatbots: A Case Study of Replika. http://hdl.handle.net/10125/79590

Yang, Y., Liu, Y., Lv, X., Ai, J., & Li, Y. (2022). Anthropomorphism and customers’ willingness to use artificial intelligence service agents. Journal of Hospitality Marketing & Management, 31(1), 1–23. https://doi.org/10.1080/19368623.2021.1926037

Yee, N., Bailenson, J. N., & Ducheneaut, N. (2009). The Proteus Effect. Communication Research, 36(2), 285–312. https://doi.org/10.1177/0093650208330254

---