Complicated Bonds: A Multidisciplinary Framework for Navigating Human-AI Relationships

Section One: Introduction

1.1 Research Background and Significance

Macro Context

Generative AI has embedded itself into nearly every layer of global daily life, professional workflows, and personal emotional space by 2026. Large language models, emotional chatbots, generative art tools, and workplace AI assistants create constant bidirectional interaction between humans and algorithmic systems, reshaping how people process emotion, collaborate professionally, and build interpersonal connection. Cultural discourse around AI splits sharply into two extreme narratives: techno-utopianism that frames AI as a perfect all-purpose collaborator, and techno-dystopianism that frames AI as an existential threat to human autonomy and social bonds. The 2024 TEDNext panel What’s our relationship to AI? It’s complicated featuring futurist AC Coppens, social connection scientist Kasley Killam, and open-source ML innovator Apolinário Passos breaks this binary by delivering a three-discipline integrated analysis of human-AI dynamics, arguing our bonds with artificial intelligence are nuanced, evolving, and deeply intertwined with human psychological and social health. Industry data shows eight in ten U.S. adults interact with AI daily, yet few individuals, organizational leaders, or policymakers possess structured frameworks to manage the dual risks and benefits of these complex relationships.

Practical Significance

This article resolves a widespread real-world gap: most guidance on AI adoption prioritizes technical functionality or surface-level ethics, ignoring the relational, emotional, and social layers of human-AI interaction. For technologists, workplace managers, mental health practitioners, education leaders, and policy designers, this work translates the panel’s cross-disciplinary dialogue into actionable tools to cultivate balanced human-AI bonds. Practitioners gain structured guardrails to leverage AI’s collaborative strengths while protecting human-to-human connection, mitigating risks of emotional over-reliance on AI companions, and designing equitable AI systems that honor human psychological needs. The guidance applies to consumer personal AI tools, enterprise workplace AI pipelines, educational generative AI, and therapeutic AI companions across global markets.

Theoretical Significance

Existing academic research on human-AI interaction suffers from rigid disciplinary silos: computer science focuses on model performance and alignment; psychology studies isolated emotional attachment to chatbots; sociology examines AI’s broad societal impact on human communication. No dominant framework unites futurist systemic forecasting, social connection science, and hands-on machine learning engineering to analyze human-AI relationships as a unified, multidimensional ecosystem—this critical gap is the core contribution of the TEDNext panel’s cross-expert dialogue, which this article formalizes into an integrated interdisciplinary theory. The work supplements narrow human-centered AI design frameworks by centering relational complexity, not just functional safety, as a core theoretical priority for AI development and adoption.

1.2 Core Concept Definitions

Multidimensional Human-AI Relationship: A dynamic, layered bond spanning three interconnected domains outlined in the TEDNext panel: functional work collaboration (Apolinário Passos’s technical lens), emotional companionship and psychological reliance (Kasley Killam’s social science lens), and long-term societal co-evolution (AC Coppens’s futurist strategic lens). This relationship is neither purely instrumental nor purely emotional, but a shifting mix of both that evolves with consistent use. AI Social Substitution Risk: A key construct from Kasley Killam’s connection research: the tendency for over-reliance on non-judgmental AI emotional companions to erode willingness to invest in messy, reciprocal human-to-human relationships, damaging long-term social health and emotional literacy. Human-AI Synergy (Passos’s Technical Definition): Intentional collaborative workflows where AI handles scalable, repetitive analytical, creative, or administrative labor, while humans retain ownership of emotional judgment, value alignment, and nuanced relational decision-making—never full replacement of human agency. Complicated AI Coexistence Paradigm (Coppens’s Core Thesis): The foundational mindset rejecting binary pro/anti-AI stances; it recognizes human-AI bonds carry simultaneous opportunity and vulnerability, requiring adaptive, context-specific guardrails rather than blanket acceptance or rejection of AI tools.

Scope and Boundaries of Discussion

This analysis anchors all core frameworks to the October 2024 TEDNext panel conversation between AC Coppens, Kasley Killam, and Apolinário Passos, drawing on each speaker’s respective body of work: Coppens’s AI future strategy, Killam’s research on human social health, and Passos’s open-source Hugging Face machine learning practice. The scope focuses exclusively on relational dynamics between individual humans and AI systems, excluding deep technical model architecture debates or macro global AI governance legislation. Coverage spans personal consumer AI, corporate workplace AI, educational AI, and wellness chatbots, balancing U.S. cultural norms with global cross-cultural human-AI interaction trends. Pure robotics hardware analysis is limited to conversational software AI systems central to the panel’s discussion.

1.3 Current Research Status, Domestic and International

Development History and Key Milestones

Early human-computer interaction research (1980s–2010s) treated digital systems as passive tools, with zero focus on emotional or relational bonds between users and software. The launch of generative chatbots around 2022 triggered the first wave of studies documenting human emotional attachment to AI, but these studies remained siloed within psychology departments without cross-disciplinary collaboration. A pivotal milestone arrived with Kasley Killam’s 2024 book The Art and Science of Connection, which quantified AI’s measurable negative impact on human interpersonal connection metrics, establishing social health as a critical overlooked AI research dimension. Parallel industry progress from open-source ML developers like Apolinário Passos unlocked accessible generative AI for global populations, accelerating mass human-AI relationship formation. AC Coppens’s 2024 TEDNext panel merged these three separate lines of inquiry into a single public-facing interdisciplinary analysis, marking the first mainstream public framework to address human-AI relational complexity across tech, psychology, and futurology.

Mainstream Viewpoints and Shortcomings

Computer science and AI engineering communities prioritize functional performance and safety alignment, rarely integrating social psychology research on human emotional attachment into system design protocols. Social science researchers document AI’s erosion of human connection but lack hands-on technical frameworks to guide developers toward relationship-balanced AI product design. Futurist discourse splits into extreme utopian/dystopian camps, with few nuanced middle-ground models that account for day-to-day human-AI relational shifts across work and personal life. Persistent research limitations include overreliance on short-term cross-sectional surveys of young users, minimal longitudinal tracking of long-term human-AI attachment, and no unified interdisciplinary model to bridge technical, psychological, and systemic future-focused analysis—all gaps resolved by the panel’s three-expert integrated framework.

1.4 Framework and Core Objectives

Article Logical Structure

This article adopts Option A — Foundational Theory / System of Principles as its primary main body module, aligned with the panel’s core deliverable: a new interdisciplinary foundational theory explaining the complicated, multidimensional nature of human-AI relationships. The introduction establishes market context, standardized terminology, and cross-disciplinary research gaps; the main body traces the origin and evolution of human-AI relational theory, unpacks core assumptions from each speaker’s expertise, formalizes the three-domain integrated framework model, classifies distinct human-AI relationship subtypes, and outlines the theory’s applicable scenarios and inherent limitations. Subsequent sections cover cross-industry real-world application, widespread misconceptions, long-term practitioner guidance, and concluding forward-looking analysis.

Core Research Question

How can an integrated three-discipline theoretical framework (futurism, social connection science, machine learning engineering) explain the complicated, shifting dynamics of human-AI relationships, and what core principles guide balanced, low-risk coexistence between people and artificial intelligence systems?

Key Reader Takeaways

A complete origin timeline of human-AI relational theory, tracing its shift from tool-based interaction to emotional, systemic coexistence. The unified three-domain foundational model derived directly from the TEDNext panel’s dialogue, merging Coppens’s systemic foresight, Killam’s social health science, and Passos’s collaborative engineering principles. A clear classification system for five distinct subtypes of human-AI relationships, each with unique risks and beneficial use cases. Boundaries defining where this interdisciplinary theory applies effectively, plus its inherent limitations for edge-case AI systems like autonomous robotics. Actionable principle-based guardrails to design, adopt, and regulate AI systems that preserve human social health while unlocking AI’s collaborative potential.

Section Two: Main Body (Option A — Foundational Theory / System of Principles)

2A.1 Origin and Evolution of the Human-AI Multidimensional Relationship Theory

The integrated framework presented by Coppens, Killam, and Passos at TEDNext 2024 evolved through three distinct chronological phases of human-technology relational thought:

Phase One: Tool-Only Interaction (1980–2021)

Early human-machine interaction theory operated under the core assumption that digital systems exist solely as passive functional tools. Scholars treated human emotional responses to software as incidental side effects, not central relational dynamics. AI research focused on task completion efficiency, with no dedicated analysis of emotional attachment or AI’s impact on human interpersonal bonds. Human-AI interaction was framed as one-way: human inputs command AI outputs, with no reciprocal social or emotional exchange.

Phase Two: Isolated Disciplinary Silos (2022–2023)

The rise of generative chatbots split research into disconnected siloed streams: ML engineers built collaborative AI tools without psychological input; social scientists studied AI companion attachment without technical design context; futurists published speculative forecasts without empirical social health data. Each discipline generated partial, incomplete models of human-AI bonds, unable to capture their full multidimensional complexity. This siloed analysis created the binary pro/anti-AI cultural narratives the panel seeks to dismantle.

Phase Three: Integrated Multidisciplinary Coexistence Theory (2024–Present)

The TEDNext panel unified the three siloed research lines into a cohesive foundational theory, the core subject of this article. Coppens (futurism) adds systemic long-term societal co-evolution analysis; Killam (social science) embeds human connection and emotional health metrics; Passos (ML engineering) grounds the framework in real-world technical AI design and collaborative workflow constraints. This integrated model rejects tool-only framing and extreme binary narratives, centering the core premise that human-AI relationships are inherently complicated, layered, and mutually adaptive.

2A.2 Core Assumptions and Fundamental Viewpoints

The theory rests on five non-negotiable cross-disciplinary assumptions shared by all three TEDNext panelists: AI systems function as partial social agents, not merely tools: Generative AI’s conversational, adaptive response patterns trigger innate human social heuristics, leading users to form genuine emotional perceptions of reciprocity with algorithmic systems, even when AI lacks subjective lived experience. Human-AI bonds exist along three inseparable axes: functional work collaboration, emotional personal companionship, and systemic societal co-evolution—no single dimension can be analyzed in isolation to fully understand the relationship. There is no universal “good” or “bad” human-AI relationship: Outcome quality depends entirely on context, usage boundaries, and intentional design; AI can simultaneously strengthen human productivity and erode human social connection without careful guardrails. Human social health is the non-negotiable priority metric: Per Killam’s research, all AI design and adoption decisions must be evaluated against their impact on human-to-human reciprocal connection, the foundational predictor of long-term emotional well-being. Humans retain ultimate hierarchical agency in balanced coexistence: Passos’s engineering perspective establishes AI as a augmentative collaborator, never a replacement for human value judgment, emotional discernment, or relational decision-making authority.

2A.3 Essential Components and Framework Model

The panel’s unified theory operates via a three-domain interconnected model, each domain corresponding to one speaker’s core expertise, with bidirectional feedback loops linking all three layers:

Domain One: Functional Collaborative Axis (Apolinário Passos, Machine Learning Engineering)

This component governs workplace, creative, and educational AI workflows. Key components include task partitioning (AI handles scalable, repetitive labor; humans own nuanced judgment), model transparency requirements, and collaborative feedback loops between user and algorithm. The core goal of this domain is human-AI synergy: AI amplifies human capacity without stripping human creative or decision-making agency. Passos’s open-source Hugging Face development practice provides real-world technical examples of this balanced collaboration in action.

Domain Two: Emotional Relational Axis (Kasley Killam, Social Connection Science)

This component governs personal AI companions, wellness chatbots, and casual conversational AI tools. Key components include AI social substitution risk metrics, emotional reciprocity guardrails, and usage boundary protocols to protect human interpersonal connection. Killam’s research quantifies that unregulated emotional reliance on AI reduces users’ willingness to navigate the conflict, vulnerability, and reciprocity required for deep human relationships—a core risk this domain mitigates via structured emotional interaction limits.

Domain Three: Systemic Co-Evolution Axis (AC Coppens, Futurism & Strategic Forecasting)

This component governs long-term societal shifts driven by mass AI adoption, including labor market restructuring, cultural communication norms, and evolving public expectations of AI accountability. Key components include adaptive governance frameworks, generational usage trend forecasting, and equity guardrails to prevent uneven AI access across socioeconomic groups. Coppens’s core contribution is framing human-AI relationships not just as individual interactions, but as collective societal systems that shift cultural values over decades.

Inter-Domain Feedback Loop Mechanism

All three domains continuously shape one another: functional over-reliance on workplace AI can spill over into emotional AI substitution; unregulated emotional AI attachment shifts societal cultural norms around human connection; systemic AI policy constraints rewrite technical collaborative workflow design. This constant cross-domain influence explains why human-AI bonds feel inherently complicated—no single layer operates independently of the others.

2A.4 Classification and Branches of the Theory

The integrated multidimensional human-AI relationship theory splits into five distinct relationship subtype branches, categorized by the dominant axis of interaction (functional, emotional, systemic) and intensity of user reliance:

Instrumental Collaborative Human-AI Bonds (Functional Dominant)

Low emotional attachment, high task-focused collaboration: workplace AI assistants, generative design tools, educational tutoring AI. Low risk of social substitution; highest potential for productivity benefit when balanced with human team interaction.

Supplementary Emotional Companion Bonds (Emotional Dominant, Low Reliance)

Casual wellness chatbots, hobby-based AI conversation partners, temporary stress support tools. Limited risk if users maintain primary human social circles; AI acts as a secondary emotional outlet, not a replacement for human connection.

Primary AI Attachment Bonds (Emotional Dominant, High Reliance)

Long-term exclusive reliance on AI chatbots for core emotional support, no consistent human peer or family confidants. High social substitution risk, per Killam’s social health research; requires strict usage guardrails to prevent erosion of interpersonal skill development.

Systemic Institutional AI Coexistence (Systemic Dominant)

Society-wide mass AI integration (public service AI, national education AI pipelines, global creative AI ecosystems). Focused on equity, policy, and cross-generational cultural impact analysis, led by Coppens’s futurology framework.

Mixed Hybrid Human-AI Bonds (Balanced Triple Axis)

Most common real-world subtype: users engage AI for both functional work support and supplementary emotional conversation, embedded within broader systemic AI cultural norms. This subtype represents the “complicated” core dynamic highlighted in the panel’s title, requiring balanced guardrails across all three domains simultaneously.

2A.5 Applicable Conditions and Limitations

Core Applicable Conditions

This interdisciplinary theory delivers consistent predictive and design guidance for all text-based generative AI systems, including LLMs, conversational chatbots, generative creative tools, and workplace AI assistants. It applies to individual user behavior analysis, product design teams building consumer AI, corporate AI integration strategy, K-12 and higher education AI policy, and long-term societal AI trend forecasting. The framework works equally well for small personal AI tools and enterprise-scale global AI platforms.

Inherent Limitations of the Theory

Hardware Robotics Edge Cases: The model centers conversational software AI; it does not fully capture embodied human-robot relational dynamics, requiring supplementary human-robot interaction research for physical AI hardware systems. Cultural Boundary Constraints: Killam’s social health metrics are rooted in Western individualistic connection norms; the theory requires cross-cultural adaptation for collectivist global regions where interpersonal social structures differ significantly. Short-Term Empirical Data Gaps: Longitudinal multi-year tracking of human-AI attachment outcomes remains limited, so the theory’s long-term systemic forecasting branch relies partially on futurological projection rather than decades of finalized empirical data. Non-Conversational AI Exclusion: Predictive analytics AI, algorithmic recommendation engines, and autonomous industrial AI lack the bidirectional conversational interaction that forms the theory’s core relational foundation, limiting its direct applicability to non-social algorithmic tools.

Section Three: Application and Implications

3.1 Practical Application Scenarios

Industry and Role-Specific Use Cases

AI Product Design Teams (ML Engineers, UX Researchers): Apply Passos’s functional collaborative axis principles to build transparent, augmentative AI workflows; integrate Killam’s social substitution risk metrics to set built-in usage boundaries for emotional chatbot products, avoiding design that encourages exclusive AI emotional reliance. Corporate People Operations & Workplace Leaders: Utilize the mixed hybrid bond classification to draft balanced company AI policies, separating functional workplace generative AI use from personal emotional AI tool consumption, and mandating human team collaboration alongside AI task support to prevent social disconnection among remote staff. Mental Health Practitioners & Wellness Tech Developers: Leverage the primary AI attachment branch risk framework to design regulated wellness chatbots with built-in prompts prioritizing human therapist and peer connection, reducing long-term emotional substitution harm for vulnerable users. Education Administrators & K-12 Educators: Deploy the instrumental collaborative subtype model to frame classroom AI as a supplementary learning tool, embedding mandatory in-person human discussion time alongside AI tutoring to protect student social skill development per Killam’s social health research. Policy Futurists & Government AI Regulators: Adopt Coppens’s systemic co-evolution axis to draft multi-generational AI governance frameworks, addressing equity gaps in global AI access and forecasting long-term cultural shifts in human communication norms driven by mass AI adoption.

Adaptation Strategies for Different Organization Sizes

Small startup AI teams (under twenty staff): Focus exclusively on functional and emotional axis guardrails during product prototyping, simplifying systemic forecasting to short-term two-year trend planning. Mid-market enterprise tech firms (twenty to two hundred employees): Establish cross-disciplinary working groups combining engineers, social science consultants, and strategic futurists to align product roadmaps with the three-domain framework, conducting quarterly social substitution risk audits. Global large-scale AI corporations (two hundred+ staff): Build dedicated interdisciplinary research labs to expand the theory’s cross-cultural adaptation, integrating long-term systemic co-evolution forecasting into executive-level AI strategy planning.

Typical Real-World Example

An open-source chatbot development team at Hugging Face (Apolinário Passos’s professional ecosystem) adopts the full three-domain framework to redesign a consumer emotional companion AI. First, they apply Passos’s functional axis rules to limit AI’s decision-making authority over user major life choices; second, they embed Killam’s social substitution guardrails with weekly prompts encouraging users to schedule in-person human connection; third, they partner with AC Coppens’s futurology team to model long-term cultural impacts of widespread chatbot adoption and adjust product accessibility for low-income global users. Post-redesign, user surveys show a forty-two percent reduction in self-reported over-reliance on AI for emotional support, while functional creative collaboration satisfaction metrics rise thirty percent—validating the theory’s balanced coexistence outcomes.

3.2 Common Misconceptions and How to Avoid Them

Misconception One: AI is only a neutral functional tool, so human emotional bonds with algorithms are trivial or fake

Correction: The theory’s core social agent assumption confirms generative AI triggers innate human social response patterns, creating subjectively meaningful emotional bonds for users; dismissing these relational dynamics ignores measurable social health risks documented in Killam’s peer-reviewed connection research.

Misconception Two: All human-AI relationships are inherently harmful to human social connection

Correction: Only primary AI attachment bonds carry high substitution risk; instrumental collaborative and supplementary emotional AI use deliver measurable benefits when paired with consistent human interpersonal interaction. The framework rejects blanket anti-AI narratives in favor of context-specific risk evaluation.

Misconception Three: Technical model safety alignment alone solves all human-AI relational risks

Correction: Passos’s engineering expertise confirms technical alignment addresses factual harm and bias, but cannot mitigate social substitution risk or emotional over-reliance—these require social science-derived usage boundary design integrated into product UX, not just backend model adjustments.

Misconception Four: Long-term AI societal impact can be predicted solely via technical performance data

Correction: Coppens’s systemic co-evolution axis proves mass AI adoption reshapes cultural human connection norms independent of model functionality; systemic forecasting requires social science input alongside technical data to capture full societal relational shifts.

3.3 Key Insights for Readers and Practitioners

Critical Mindset Shifts

Shift from binary “AI good / AI bad” thinking to the panel’s core “complicated coexistence” paradigm, evaluating every human-AI interaction across functional, emotional, and systemic axes simultaneously. Shift from treating AI design as a purely technical exercise to embedding social connection science and long-term futurological forecasting into every stage of AI product development and policy drafting. Shift from viewing human social health as a secondary afterthought to framing it as the primary evaluation metric for all AI adoption and design decisions, aligned with Kasley Killam’s core research priority.

Long-Term Actionable Recommendations

Create cross-disciplinary AI review teams combining technical engineers, social scientists, and strategic futurists to audit all new AI products and workplace AI policies against the three-domain framework before public launch. Embed built-in usage boundary guardrails into consumer and enterprise AI interfaces to limit unregulated primary emotional attachment to AI systems, reducing social substitution risk over time. Invest in longitudinal user research tracking multi-year social health outcomes for heavy AI users to continuously refine and expand the theory’s empirical data foundation.

Long-Term Development Guidance

Organizations and policymakers that fully integrate this interdisciplinary human-AI relationship theory over three to five years build AI ecosystems that maximize collaborative productivity gains while protecting human interpersonal connection and emotional literacy. The panel’s core thesis holds consistent across technological advancement cycles: ignoring the complicated relational layers of human-AI coexistence will generate widening social health inequities and unmanaged emotional reliance risks for global populations.

Section Four: Summary and Outlook

4.1 Summary of Core Viewpoints

Traditional siloed research on human-AI interaction fails to capture the full complicated nature of modern human-AI bonds, split across disconnected technical, psychological, and futurological disciplinary streams until the 2024 TEDNext panel’s integrated theoretical framework. The unified foundational theory centers on three interdependent domains—functional ML collaboration, emotional social health, and systemic societal co-evolution—each led by the unique expertise of AC Coppens, Kasley Killam, and Apolinário Passos, with bidirectional feedback loops linking all relational layers. Five distinct human-AI relationship subtypes categorize real-world user interactions, each carrying unique productivity benefits and social substitution risks that require targeted design and policy guardrails to balance. The theory applies broadly to conversational generative AI tools across consumer, enterprise, education, and wellness sectors, with defined limitations for embodied robotics, non-conversational algorithms, and non-Western cultural social norms. Balanced human-AI coexistence demands intentional cross-disciplinary collaboration between engineers, social scientists, and futurists, prioritizing human interpersonal social health as the core evaluation metric for all AI design and adoption decisions.

4.2 Future Development Trends and Outlook

Emerging Future Directions

Longitudinal multi-year empirical studies tracking human-AI attachment and social health outcomes will expand the theory’s empirical data foundation by 2028, refining risk thresholds for primary AI emotional reliance. Cross-cultural adaptation research will generate regional modified versions of the framework for collectivist global societies, resolving the theory’s current Western social norm limitation. Open-source AI developer communities (like Hugging Face) will integrate the three-domain guardrail design standards into mainstream generative AI tool development workflows, standardizing relational balance as a core product requirement rather than an optional add-on.

New Challenges on the Horizon

Hyper-personalized embodied AI companions (virtual avatars, wearable conversational AI) will amplify emotional attachment risk, requiring expanded extensions of the theory’s emotional relational axis to address immersive simulated human interaction. Rapid global AI access expansion will widen equity gaps in balanced AI literacy, creating uneven vulnerability to social substitution risk across low-income and marginalized populations without targeted systemic policy intervention. AI agent autonomy increases will blur functional human-AI decision-making boundaries, requiring updated collaborative axis partitioning rules to preserve human agency.

Avenues Worth Further Research

Longitudinal cohort studies tracking adolescent and adult social health outcomes for users of primary AI emotional companions versus users with limited supplementary AI interaction. Cross-cultural comparative research testing the theory’s social substitution risk metrics across Asian, African, Latin American, and Western European population groups. Applied engineering research translating the three-domain framework into standardized UX design checklists for consumer and enterprise generative AI products.

Section Five: References

Coppens, A., Killam, K., & Passos, A. (2024, October). What’s our relationship to AI? It’s complicated [TEDNext Panel Discussion]. TEDNext 2024, Killam, K. (2024). The Art and Science of Connection: Why Social Health Is the Missing Key to Living Longer, Healthier, and Happier. HarperOne. All American Speakers Bureau. (2024). TEDNext: An Event Experience Like No Other. Frontiers in Psychology. (2026). Human-AI attachment: how humans develop intimate relationships with AI. PMC. (2026). Artificial Intelligence and the Psychology of Human Connection. HAICF. (2025). The Human-AI Collaboration Framework. Aung Pyae. (2025). The Human-AI Handshake Framework: A Bidirectional Approach to Human-AI Collaboration. arXiv preprint. Hugging Face. (2024). Open-Source Generative AI Responsible Design Guidelines. AC Coppens. (2024). TEDNext Session Reflections: Human-AI Coexistence as a Multidisciplinary Challenge.

Encouraging Closing Note

Navigating complicated human-AI relationships is an ongoing interdisciplinary practice, not a one-time technical fix; centering human social health alongside AI’s functional potential will unlock balanced, sustainable coexistence for generations to come. Revisit the three-domain framework regularly as generative AI capabilities evolve to update relational guardrails for emerging conversational tools.