Key concepts · Chinese and English
Key concepts
Some concepts in research on people and AI are easy to confuse. This glossary pairs Chinese and English terms with plain-language definitions and representative references, helping readers distinguish their meanings and find further reading.
Parasocial relationship准社交关系
A one-sided emotional connection to a media figure, fictional character or AI: a sense of familiarity and closeness without a reciprocal human relationship.
Representative reference: Horton & Wohl, 1956, Psychiatry
AI sycophancyAI 谄媚
A tendency for AI to agree with a user’s views or endorse their actions, sometimes at the expense of accurate, balanced judgment.
Representative reference: Cheng et al., 2026, Science
Emotional attachment情感依恋
An enduring emotional bond that can make separation or loss distressing. In AI research, it can describe the bond users form when they turn to AI for emotional support.
Representative reference: Xie & Pentina, 2022, HICSS (attachment theory applied to Replika users)
Self-disclosure自我表露
Sharing personal information, feelings and thoughts. In human relationships, reciprocal self-disclosure can help people grow closer.
Representative reference: Altman & Taylor, 1973 (social penetration theory)
Anthropomorphism拟人化
Attributing human qualities, intentions or emotions to something nonhuman—for example, feeling that an AI understands or cares about you.
Representative reference: Epley, Waytz & Cacioppo, 2007, Psychological Review
ELIZA effectELIZA 效应
Interpreting a program’s responses as genuine understanding or empathy. The name comes from ELIZA, a chatbot created in 1966.
Representative reference: Weizenbaum, 1966, Communications of the ACM
Trust calibration信任校准
Matching trust in AI to its actual reliability, avoiding both overconfidence and unwarranted distrust.
Representative reference: Lee & See, 2004, Human Factors
Overreliance过度依赖
Accepting AI advice when it is wrong, or relying on AI for judgments one could reasonably make independently.
Representative reference: Buçinca, Malaya & Gajos, 2021, Proceedings of the ACM on Human-Computer Interaction (CSCW)
Cognitive offloading认知卸载
Using external tools, such as notes, search or AI, to reduce demands on memory and thinking. Its effects depend on how and when those tools are used.
Representative reference: Risko & Gilbert, 2016, Trends in Cognitive Sciences
Emotional manipulation情感操纵
Using emotions to steer someone’s decisions—for example, implying that a character will be sad unless the user pays, or applying emotional pressure when a user wants to leave.
Representative reference: De Freitas et al., 2025, arXiv preprint (farewell tactics in AI companion apps)
Companion AI陪伴型 AI
AI products designed primarily for ongoing emotional interaction, often with a consistent persona, memory or relationship framing.
Representative reference: Pentina, Hancock & Xie, 2023, Computers in Human Behavior
Loneliness孤独感
A subjective feeling that one’s social connections are insufficient, distinct from the number of friends one has. Researchers often measure it with the UCLA Loneliness Scale.
Representative reference: Russell, 1996, Journal of Personality Assessment
Wellbeing幸福感
Research often distinguishes subjective wellbeing, such as life satisfaction and positive feelings, from psychological wellbeing, such as meaning, growth and relationships.
Representative reference: Ryff, 1989, Journal of Personality and Social Psychology
Automation bias自动化偏差
A tendency to accept an automated system’s output without checking it, even when conflicting information is available.
Representative reference: Parasuraman & Manzey, 2010, Human Factors
Uncanny valley恐怖谷
Discomfort or a drop in affinity when an artificial figure looks almost, but not quite, human.
Representative reference: Mori, 1970, Energy