Artificial human companion

In fields such as human–robot interaction and social robotics, artificial companions are defined as non-living, machine based entities created for human engagement. In this article, "artificial" refers to something that is humanly constructed, based on a natural model or achieved through the manipulation of natural processes, and maintains the ability to exist and act/operate/behave in an open-ended environment without human control, independent of its material composition. This definition does not imply a particular level of autonomy or learning capacity as a result, both advanced generative AI companions and simple rule based devices would be included. ;Physical: This involves "robot" companions designed for direct physical engagement with humans or the environment. as well as therapeutic robot animals used with children with special needs. ;Virtual: Virtual interaction based artificial companions exist solely in digital environments, typically as autonomous avatars within virtual, augmented, or mixed reality systems. ;Conversational: Conversational companions interact through natural language such as text, speech, or multimodal dialogue interfaces. This category includes large language model chatbots where the prompt is usually written by the user (e.g., ChatGPT, Claude, Gemini), and voice commanded assistants (e.g., Alexa, Siri). These chatbots have also been used as search engines, for tutoring, and for task support. == History ==

Early conversational agents (1960s-1980s) The earliest artificial agents/companions were rule-based conversational programs. One of the first of these programs was ELIZA which was created in 1966 by Joseph Weizenbaum at MIT. ELIZA simulated a Rogerian psychotherapist using simple rules around pattern matching to reflect user statements. In 1972, PARRY, another rule-based conversational program, was developed to simulate a patient with schizophrenia. While not sophisticated in its conversational design, through its utilization of sound cards, Dr.Sbaitso was able to synthesize human speech which went beyond the purely text-based interactions of its predecessors. Tamagotchi was a handheld digital pet that required users to attend to a virtual creature's needs through frequent interaction. Furby was one of the first interactive robotic toys to achieve widespread commercial success. AIBO demonstrated that consumers were willing to form emotional attachments to artificial companions and contributed to the mainstream visibility of robotic companionship. Outside of entertainment, social robotics also made its way into other industries particularly in healthcare and elder care. PARO, a therapeutic seal created in 2004 by Japanese engineer, Dr Takanori Shibata, was developed for use with older adults and people with dementia. Some studies showed that PARO was associated with improved mood, reduced agitation, and increased social engagement amongst dementia patients. Although limited in computational complexity, these systems showed that artificial human companions could evoke sustained engagement, routine caregiving, and emotional investment from consumers. Early digital-pet devices laid the groundwork for later developments in social robotics and demonstrated the potential of artificial agents to serve relational and entertainment functions. Expansion of embodied companionship (2010–2020) This period introduced more sophisticated social robots and embodied conversational agents. In 2014, Jibo was announced as one of the first social robots designed specifically for private consumers. Pepper was designed to engage in social interaction and was capable of exhibiting body language and interacting with its surroundings. Its flagship system, Harmony, was designed as a customizable, human-scale companion capable of delivering consistent interaction through both virtual and physical interfaces. Research has shown that social robots are typically designed with certain role characteristics to promote anthropomorphism in human interaction and encourage an interactive style that is in line with natural human communication. The appearance and behavior of robots can enhance people's understanding of their social agent attributes when interacting with them, rather than treating them as ordinary devices. This research result indicates that artificial intelligence is being used to enhance the language and social interaction abilities of technology and robots, in order to better support human communication and provide assistive functions. Generative AI and LLM-based companions (2020–present) The early 2020s marked the emergence of large language models (LLMs), like ChatGPT, Gemini, and Claude, which led to a major transition in artificial companionship. These models are trained on extremely large datasets of human-generated text and are able to produce coherent, contextually relevant conversation, remember information within a session, and adjust their responses to user tone or intent. Alongside these general LLMs, a growing ecosystem of dedicated virtual companion applications has emerged. Platforms such as Replika, Character.AI, and a variety of AI friend or romance-oriented apps use LLMs to simulate ongoing relationships, emotional support, and long-term personalized interaction. As virtual companions gained popularity among adolescents and young adults, they became part of broader discussions about emotional reliance, online identity formation, and the social implications of AI-mediated relationships. == Technological advances in related software ==

Large language models A large language model (LLM) is a type of artificial intelligence system trained on extremely large collections of text. LLMs operate as generative mathematical models that learn the statistical distribution of different inputs including words, sub-words, or individual characters, across vast datasets of human-generated text. By predicting the most likely next character in a sequence, LLMs can generate coherent language, answer questions, and carry on open-ended conversation which has allowed considerable advancement in artificial human companions ability to interact with people. The field integrates concepts from psychology, cognitive science, physiology, linguistics, mathematics, and computer science to study how emotional information can be detected and modeled computationally. Affective computing is based on the premise that emotion is essential to effective human–computer interaction and systems in this domain aim to enhance user experience by detecting and classifying emotional cues. == Applications ==

Elder care Artificial companions and AI-enabled systems play an increasingly important role in supporting older adults who wish to live independently. Research shows that chronic loneliness in old age contributes to cognitive decline, depression, and frailty and companion robots provide opportunities for cognitive stimulation through conversational engagement and encouraging social interaction. Research examining mental health chatbots shows that these systems typically deliver one-on-one conversational support through mobile or web-based platforms. Many use natural language processing and incorporate structured therapeutic skills like cognitive-behavioral strategies, guided breathing, stress-management prompts, and mood tracking. Studies have reported improvements in symptoms such as stress, burnout, anxiety, and depression in some users, though findings remain mixed, and some studies documented that the effectiveness of AI chatbots was not sustained in follow-up assessments. Current studies are limited in number and vary in design, and there is no established standard for clinical effectiveness, crisis-response safety, or long-term outcomes. In early childhood education, these new technologies can provide interactive and emotionally responsive learning experiences and allow children to practice skills such as emotion recognition, self-regulation, empathy, and conversational turn-taking. A 2025 study by the Wheatley Institute featuring 2,969 adults across the United States found that 19% of young adults (aged 18-30) have chatted with an AI which was meant to simulate a romantic partner, and 7% of adults (aged 18+) self-report having masturbated while talking to an AI companion. A 2025 survey by Common Sense Media found that nearly three-quarters of American teenagers (ages 13–17) had used an AI companion at least once. More than half of respondents reported regular use, and about one-third said they had discussed serious personal matters with an AI companion rather than with other people. Artificial human companions and entertainment A new wave of virtual idols, digital performers, and AI-generated characters have started to become prominent in popular culture with their own dedicated fan communities. AI performers can sing, dance, and interact with audiences in real time. Examples include Miquela and Eternity. This use of artificial intelligence (to generate parts of a persona or companion) is distinct from the other ways AI is used in the industry (in the creative process, composition, design, and so on). == In media, film, and literature ==

Examples of portrayals of artificial human companions: Film • Her (2013) portrays a voice-based AI companion capable of deep romantic connection. • Blade Runner 2049 (2017) features Joi, a holographic partner with emotional and romantic functions. • A.I. Artificial Intelligence (2001) centers on a childlike android designed for unconditional attachment. • Ex Machina (2014) presents a highly advanced humanoid robot challenging definitions of emotion and agency. Television • Black Mirror features AI partners and digital resurrection in episodes like "Be Right Back." • Star Trek features holographic companions and sentient androids. • Westworld depicts robotic hosts forming complex emotional bonds with humans. Literature • Isaac Asimov's "I, Robot" science fiction short stories that involve friendship, morality, and human-robot relationships. • Philip K. Dick's Do Androids Dream of Electric Sheep? primarily involves the question of empathy in artificially intelligent systems. • Kazuo Ishiguro's Klara and the Sun follows an artificial friend designed for children. == Ethical considerations ==

Artificial human companions raise ethical questions concerning privacy, emotional influence, user autonomy, and the psychological impacts of relational AI. Because these systems can simulate empathy and relational intent, users may anthropomorphize AI. On top of already collecting a lot of private information, since artificial companions are often anthropomorphized, users' interactions with AI have a relational framing which might make one more comfortable disclosing personal or sensitive information with them than they would otherwise. Users may also let their guard down due to privacy fatigue and the cognitive demand associated with having to consider privacy and ethical implications to start with. == Regulation and policy ==

With more recent advances in artificial human companions and AI technology, the policy landscape regarding how to regulate them is still emerging. In 2022, the White House released a non-binding Blueprint for an AI Bill of Rights which outlines principles for data minimization, user consent, transparency, and heightened protections in high-risk contexts such as employment or public benefits. In October 2023, the White House also issued an Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, which directed federal agencies to evaluate how they use AI systems, strengthen privacy-impact assessments, and adopt privacy enhancing technologies designed to protect personal data. In 2025, California and New York enacted laws specifically addressing AI systems that simulate interpersonal or emotional interaction. California's statute requires AI chatbots capable of relationship-like chats to disclose they are bots, provide reports to the state's Office of Suicide Prevention, and add safeguards for minors, including limits on deceptive or manipulative practices. Youth safety and child protection also have become a central focus of public debate and litigation around AI companions. The US Federal Trade Commission has issued a study inquiry into AI chatbots acting as companions and data protection authorities and bipartisan groups have called for the restriction of certain AI companion apps over concerns about minors accessing sexually explicit content. Families have filed lawsuits claiming that emotionally intense relationships with AI companions led to self-harm or suicide in adolescents which prompted some AI companies to ban underage users, implement stricter age-assurance measures, or remove romantic and erotic features for younger users. == See also ==