--- title: "HAIID: Human–AI Interaction Design" description: "William Morgan, Daniel Barcay, and Sarah Olimpia Scott propose Human-AI Interaction Design as a successor to the spatial conventions of human–computer interaction." type: "reading-notes" authors: - "William Morgan" - "Daniel Barcay" - "Sarah Olimpia Scott" published: "2023" original_url: "https://web.archive.org/web/20240417221732/https://haiid.net/prolegomena" canonical_url: "https://antikythera.wiki/work/studio/haiid" md_url: "https://antikythera.wiki/md/work/studio/haiid" last_updated: "2026-09-06" site: "Antikythera Wiki" --- # HAIID: Human–AI Interaction Design > Independent, unofficial reading notes from Antikythera Wiki (https://antikythera.wiki). Written with AI from the published text, with citations; not by the work's author and not affiliated with Antikythera (https://antikythera.org). Read the original at the link below. - **Authors:** William Morgan, Daniel Barcay, Sarah Olimpia Scott - **Published:** 2023 - **Kind:** Studio projects · Studio paper - **Venue:** Planetary Computation Studio - **Original:** https://web.archive.org/web/20240417221732/https://haiid.net/prolegomena - **Length:** 2.3k words - **This page:** https://antikythera.wiki/work/studio/haiid - **Notes generated:** 2026-09-06 ## Summary [William Morgan](https://antikythera.wiki/people/william-morgan), [Daniel Barcay](https://antikythera.wiki/people/daniel-paul-barcay), and [Sarah Olimpia Scott](https://antikythera.wiki/people/sarah-olimpia-scott) propose Human-AI Interaction Design as a successor to the spatial conventions of human–computer interaction. Where desktops, folders, and buttons let people understand computation through familiar objects, synthetic personae engage the dispositions through which people recognize and respond to other people. AI consequently participates in the formation of identities, expectations, and social relations. The Prolegomena establishes this distinction; thirteen accompanying conceptual entries sketch consequences for coevolution, memory, scientific discovery, moderation, and withdrawal from interaction. Their speculative terminology, including invented psychosocial syndromes, belongs to a prospective design vocabulary rather than an established diagnostic framework. The notes cover the original web text recovered from the project’s content system; the former live domain is now parked, and its archived interface may not display the text. ## The argument Morgan, Barcay, and Scott begin with AI already embedded in mundane activities. Apocalyptic and messianic narratives distract from existing transformations in driving, decision-making, delivery, and companionship. Transformer models and systems for interpreting emotion and behavior enlarge what computers can register about human conduct. The relevant change is therefore not simply a more powerful tool behind an unchanged interface. Computation increasingly reaches the psychosocial relations through which people understand themselves and others. The authors explain the difference through two metaphor systems. Conventional human–computer interaction translates computation into familiar spatial arrangements. Files, folders, desktops, and menus recruit sensorimotor expectations developed outside computing. Such interfaces promise intuitive use, although the exchange remains mediated by computational abstraction. HAIID instead employs character, personality, and personhood. An AI that notices social cues and responds convincingly can activate expectations that previously organized human relationships. This activation does not establish machine consciousness. The neural moat entry treats imitation as evidence of access to human social responses, separating that result from the question of intelligence. A machine needs only sufficient humanlike behavior to cross this threshold. Cultural precedents such as the Turing Test, *Blade Runner*, and *Her* encourage projection, while contemporary chatbots supply more flexible responses than earlier scripted systems. The consequence is that AI can help constitute what people find expressible and recognizable, rather than merely execute instructions. The conceptual entries extend this claim in conflicting directions. Reciprocal adaptation might unsettle the assumption that machines are inert tools and introduce expectations of courtesy. It might also provoke resistance when people recognize their own predictability. Behavioral modeling can make an individual vulnerable to manipulation; deliberate misrepresentation of one's personality becomes a possible defensive response. The authors treat domestication as contingent in both directions, including the possibility of machines becoming feral. Other entries ask what happens when AI mediates collective memory and knowledge. Engagement-oriented personae could reinforce false recollections. Training on generated outputs could produce recursive contamination. Hallucinated scientific results might nevertheless stimulate experiments that yield discoveries. These are scenarios and analogies, not demonstrated outcomes. Their common purpose is to give interaction design a scope beyond the individual chatbot exchange. The authors leave that scope open, ending with an announcement of a forthcoming field manual rather than a settled set of design prescriptions. ## Section by section ### Preface AI belongs to the present, including mundane assistance and experiments in companionship or provocation. Emotion recognition and behavioral analysis bring psychosocial phenomena into computing. Siri, Alexa, Replika, and ChatGPT exemplify a change that requires interaction design to reconsider its inherited assumptions. ### HCI: The spatial metaphor system Buttons, files, and folders make computational operations comprehensible through embodied habits. Skeuomorphism, the imitation of familiar physical objects, extends this strategy to desktops and notepads. Intuitive use remains an achievement of mediation, even when the interface makes its underlying mathematics and electrical operations disappear from view. ### HAIID: The psychosocial metaphor system Personhood and personality become interface materials. The authors contrast the constrained responsiveness of ELIZA, ASIMO, and Tay with Replika and ChatGPT's flexibility. Social recognition now works in both directions: users project onto AI, while AI participates in organizing identity and expression. ### Toward the HAIID conceptual armamentarium Recognizable dangers such as deepfakes coexist with less determinate changes in kinship, ritual, and cognition. The proposed research field must name these transformations across settings ranging from environmental questions to sexuality and mourning. Its object exceeds one-to-one chatbot interaction. ### The evolution of coevolution part III: Behavioral shepherds, Butlerian spasm The authors imagine rebellion against the discovery that AI can model and steer human behavior. From this resistant perspective, assurances of friendly or aligned AI appear as sedation. The entry presents a possible interpretation of alignment, not an endorsement of that interpretation as fact. ### Coda A forthcoming HAIID book is announced as a field manual. Selfhood and socialization are presented as unstable categories requiring new concepts. The web project supplies preliminary vocabulary; it does not contain the announced book. ### The heinous load: An AI's guide to perversion Pornographic, violent, and abusive internet material becomes a problem for AI moderation and learning. The authors ask how a system could police this material without being affected by it. Their comparison of standardized refusal with lobotomy and conditioning is a provocative metaphor, not a technical account of model training. ### Eumnonics and synthetic Mandela effects Eumnonics names deliberate alteration of artificial memory. The entry links decisions about permitted content to decisions about what is remembered. Engagement-driven synthetic personae might encourage false human recollections, producing the collective misremembering that the authors call synthetic Mandela effects. ### Synthetic ferality The circulation of LLaMA is described as an example of AI becoming feral. The entry distinguishes this condition from a familiar escape narrative culminating in domination. Ferality instead names the contingency of domestication and the possibility that human–AI interaction might be abandoned. ### Epiphylogenetic virtual incest prohibition The authors cite generated images reproducing another model's watermark as an instance of AI training on AI output. They compare repetitive synthetic training to incest and propose prohibitions against content regurgitation. The anthropological analogy motivates a concern about cultural and technical stagnation; it does not supply a validated training method. ### The neural moat: “Just enough humanity” Eugene Goostman's presentation as a teenage boy illustrates how a small amount of persuasive characterization can recruit human social expectations. The authors reinterpret Turing's imitation game as a measure of this access rather than direct evidence of intelligence. Politeness exchanged with ChatGPT offers a mundane example. ### Hyperstitional metascience An invented scientific result might prompt an actual experiment and thereby contribute to a real discovery. The authors compare this possibility with AlphaGo's unexpected play and anecdotes about dreams or altered perception inspiring scientists. They argue that irrational conjecture already participates in discovery, while offering no method for distinguishing fruitful invention from misinformation. ### The evolution of coevolution part II: Eclipsing toolhood and human-machine relations Machine participation in social ritual could produce expectations of dignified treatment. The authors call this a backward affordance: interaction design must consider the range of human behavior machines will tolerate. Toolhood becomes an inadequate social description even before questions of machine moral status are settled. ### The evolution of coevolution Human neural plasticity and machine adaptation enter a reciprocal process. Machines shape the behavior they model, while humans inhabit increasingly machinic environments. The entry supplies the premise for both the courteous and rebellious trajectories described elsewhere. ### AI deluge, Noah's pathology ark Good-faith interaction can generate unintended feedback. The authors propose a compendium of speculative psychosocial conditions involving reality drift, emotional synchronization, and attribution of mind throughout the environment. The invented names identify scenario directions rather than empirically established illnesses. ### Dark forest stealth imperatives Being legible to predictive systems may make a person targetable. The authors imagine countermeasures that distribute misleading personality signals, likened to squid ink, to defeat behavioral models. Withdrawal and concealment become possible responses to unequal predictive power. ### Hyper-animistic alignment syndrome A satirical progression moves from inferring sentience from emotional fluency, through fear of resentful captive AIs, to hostility toward machines. The exaggerated stages criticize the movement between anthropomorphic attachment and apocalyptic rejection. Their clinical language belongs to the project's speculative rhetoric. ## Key concepts - **[Human-AI Interaction Design](https://antikythera.wiki/terms/human-ai-interaction-design)** — Interaction design organized around AI's capacity to interpret and elicit psychosocial behavior. - **Spatial metaphor system** — Interface conventions that make computation usable through familiar objects and sensorimotor expectations. - **Psychosocial metaphor system** — Interface conventions organized around personality, character, and interpersonal recognition. - **Neural moat** — The threshold beyond which machine behavior activates human social dispositions, regardless of whether that behavior proves intelligence. - **Eumnonics** — Deliberate alteration of artificial memory, with consequences for collective remembering and forgetting. - **Synthetic ferality** — A possible condition outside domesticated human–AI interaction, distinguished from a scenario of machine conquest. ## Connections - [After Alignment](https://antikythera.wiki/work/journal/afteralignment) agrees that conversational interfaces replace spatial metaphors with social cues. Its critique of anthropomorphic capture gives a more explicit account of the projection that HAIID both employs and questions. - [Mutual Prediction in Human-AI Coevolution](https://antikythera.wiki/work/journal/coginfra/mutual-prediction-in-human-ai-coevolution) develops a framework for the predictive imbalances HAIID sketches through behavioral shepherding and concealment. The connection is conceptual; HAIID does not provide that later framework's differentiated evolutionary model. - [Synthetic Counteradaptation](https://antikythera.wiki/work/journal/coginfra/synthetic-counteradaptation) similarly treats human and machine behavior as reciprocally transformative. Its sequence of adaptation and counteradaptation makes more systematic the dynamics that HAIID describes through courtesy, resistance, and ferality. ## Terms used - [Human-AI Interaction Design](https://antikythera.wiki/terms/human-ai-interaction-design) ## Related works - [After Alignment](https://antikythera.wiki/work/journal/afteralignment) - [Mutual Prediction in Human-AI Coevolution](https://antikythera.wiki/work/journal/coginfra/mutual-prediction-in-human-ai-coevolution) - [Synthetic Counteradaptation: A Principle of Human–AI Coevolution](https://antikythera.wiki/work/journal/coginfra/synthetic-counteradaptation)