Embodied AI raises productivity but deepens power imbalances: algorithms displace worker decision-making and extract uncompensated behavioral data, reinforcing monopoly capital and unequal access to technological dividends; coordinated institutional, organizational, and individual rights reforms are needed to rebalance human–machine production relations.

Challenges and Reconstruction of Human-Machine Collaboration in the Era of Embodied Intelligence

Yutang Guan · June 15, 2026 · Lecture Notes in Education Psychology and Public Media

openalex theoretical n/a evidence 7/10 relevance DOI Source PDF

Using Marx's reproduction framework, the paper argues that embodied AI shifts the human–machine relationship toward collaborative co-creation that raises productivity while producing new power asymmetries—workers are disciplined and transformed into unpaid data producers, data exploitation and capital concentration persist, and access to technological benefits is unequal.

Embodied intelligence is driving the human-machine relationship from a "human-dominated" model toward "collaborative co-creation," which, while boosting productivity, also triggers deep-seated contradictions in production relations. Drawing on Marx's theory of reproduction, this paper analyzes the dilemmas of human-machine collaboration through the four stages of production, distribution, exchange, and consumption: In the production stage, workers lose decision-making power, are disciplined by algorithms, and are alienated into becoming data producers; in the distribution phase, behavioral data unconsciously generated by workers drives algorithmic iteration yet remains excluded from the distribution system, resulting in hidden data exploitation; in the exchange phase, high barriers to entry for technology and capital foster market monopolies; and in the consumption phase, high costs lead to service stratification, making it difficult for technological dividends to benefit the general public. The root causes lie in the disruption of labor relations boundaries by the transformation of the means of production, the exclusion of implicit data labor from distribution rules, the concentration of capital driven by high industry barriers, and social structural constraints on technological dissemination. Therefore, this paper advocates for the coordinated reconstruction of production relations across three levels: macro-level institutional constraints, meso-level organizational transformation, and micro-level rights protection. Through pathways such as the recognition of data labor rights, anti-monopoly regulation, and algorithmic transparency, we aim to build an equal, just, and sustainable future of human-machine collaboration.

Summary

Main Finding

Embodied intelligence (AI with physical agency) is reshaping production relations: it weakens workers' operational and decision-making subjectivity, creates unpaid "data labor" whose value is excluded from existing distribution rules, concentrates market power in ecosystem monopolies, and produces consumption stratification. These effects together risk turning "human‑machine collaboration" into a new form of capital-dominated dependence. The paper argues that resolving these dilemmas requires coordinated institutional, organizational, and rights-based reconstruction (recognize data labor rights, strengthen anti‑monopoly and algorithmic transparency, and protect worker/consumer rights).

Key Points

Conceptual frame: uses Marx's theory of reproduction (production → distribution → exchange → consumption) to analyze embodied intelligence’s systemic impacts.
Production:
- Embodied intelligence has "physical agency" (perception → decision → execution) and thus becomes a quasi‑agent rather than a passive tool.
- Algorithms take over decision-making and rhythm control; workers shift from autonomous operators to supervised executors and “data producers.”
- Deskilling and deeper alienation occur as experiential worker knowledge is supplanted by standardized data-driven processes.
Distribution:
- Massive real‑time perceptual/behavioral data are generated during production; ownership and distributive claims over this data are ambiguous.
- Current wage systems ignore implicit “data labor,” allowing firms/platforms to appropriate subsequent gains (data surplus value).
- Example cited: JD.com’s 2026 embodied‑intelligence data collection program mobilizing hundreds of thousands to generate training hours.
Exchange:
- High technical barriers, capital intensity, data accumulation, and ecosystem lock‑in create “ecological monopolies.”
- Leading firms capture data, users, and markets, reinforcing advantages and producing unequal exchange conditions for smaller producers and laborers.
Consumption:
- High component and product costs keep embodied robots scarce and expensive—pricing out ordinary households and many service providers.
- Result is service stratification: technological dividends accrue mostly to high‑income or capital‑intensive users; inclusive scenarios (elder care, household services) are neglected.
Root causes:
- Transformation of means of labor (instrumental leap) blurs boundaries of labor relations.
- Distribution rules remain anchored to industrial‑era measures (physical/mental labor), excluding data labor.
- Institutional and market structures favor capital concentration.
- Social inequality and procurement dynamics hinder broad dissemination.
Proposed reconstruction (three levels):
- Macro: institutional constraints—data labor recognition, anti‑monopoly regulation, public policy to steer inclusive R&D and procurement.
- Meso: organizational transformation—new enterprise protocols, algorithmic transparency and governance, data co‑operatives or shared asset models.
- Micro: rights protection—workers’ data/algorithmic rights, participation in governance, compensation for data contributions.

Data & Methods

Methodological approach: qualitative political‑economy critique drawing on Marx’s reproduction framework and analysis of production relations; conceptual reasoning rather than formal quantitative modeling.
Evidence and illustrative data:
- Empirical/case references used illustratively:
  - JD.com (March 2026) plan to collect millions of hours of embodied‑intelligence training data via large employee and citizen mobilization.
  - Boston Dynamics Spot price cited (~530,000 RMB) to illustrate cost barriers.
  - China per‑capita and urban disposable income (2024) used to illustrate affordability gaps.
- No original microdata analysis or econometric estimates presented; arguments rest on theory, industry examples and secondary statistics.
Limitations acknowledged by the author: lag between productivity leaps and institutional change; paper does not provide measurement methods for “data labor” value nor empirical quantification of surplus extraction.

Implications for AI Economics

Measurement and accounting:
- Need new metrics to capture the value contribution of implicit “data labor” (real‑time behavioral/perceptual data) in productivity and surplus value accounting.
- National accounts and firm-level productivity statistics may understate or misattribute gains from embodied AI; research should develop protocols for attributing value to data contributions.
Labor markets and wages:
- Embodied AI can create deskilling and displacement risks and also create unpaid data extraction; labor economics must consider non‑wage data rents and redesign compensation frameworks (data dividends, wage supplements, collective bargaining over data).
Market structure and competition policy:
- High fixed costs, data network effects, and ecosystem lock‑in justify rethinking antitrust in AI/robotics sectors (data portability, interoperability mandates, limits on ecosystem bundling).
- Policies to lower entry barriers (R&D subsidies, open standards, public datasets) can reduce ecological monopoly risks.
Redistribution and access:
- Without policy intervention, embodied AI may amplify inequality through stratified consumption; public procurement and subsidies can accelerate diffusion into inclusive use‑cases (healthcare, eldercare).
- Consider public or cooperative ownership models for shared embodied‑AI infrastructure or data commons.
Governance and institutions:
- Algorithmic transparency, auditability, and worker participatory governance are economically important for preventing efficiency‑gains from translating into concentrated rents.
- Legal recognition of data labor rights (ownership, collective bargaining over data use, compensation) could reshape firm incentives and the distribution of AI rents.
Research agenda suggestions:
- Empirical estimation of the share of output attributable to implicit data labor across sectors using embodied AI.
- Modeling distributional effects of different policy responses (data dividends, antitrust, public procurement) on welfare and inequality.
- Cost‑curve studies on embodied‑robot components to evaluate realistic timelines for price declines and diffusion scenarios.

Overall, the paper reframes embodied AI as not merely a productivity technology but as a force altering the institutional foundations of production and distribution; for AI economists, this implies urgent work on measurement, market structure, redistribution instruments, and governance designs to ensure inclusive outcomes.

Assessment

Paper Typetheoretical Evidence Strengthn/a — This is a conceptual, theory-driven paper drawing on Marxist frameworks rather than empirical data or causal inference methods, so it does not provide empirical evidence to support causal claims. Methods Rigormedium — The paper offers a structured, coherent theoretical argument across production, distribution, exchange, and consumption stages and proposes policy pathways; however, it lacks empirical operationalization, measurement, counterfactuals, and robustness checks that would be required for higher methodological rigor. SampleNo original empirical sample — a normative and conceptual analysis using Marx's theory of reproduction and qualitative reasoning about human–machine relations, algorithmic discipline, data labor, market concentration, and access to technology. Themeshuman_ai_collab labor_markets governance inequality GeneralizabilityAnalytic lens grounded in Marxist theory may not map neatly onto all institutional or cultural contexts, No empirical validation across sectors, firm sizes, countries, or specific AI technologies, Assumes relatively uniform dynamics of algorithmic discipline and data extraction that may vary by task, industry, and platform design, Policy recommendations are high-level and may not account for political feasibility or heterogeneous regulatory environments

Claims (9)

Claim	Direction	Outcome	Confidence & Evidence	Details
Embodied intelligence is driving the human-machine relationship from a "human-dominated" model toward "collaborative co-creation," which, while boosting productivity, also triggers deep-seated contradictions in production relations. Firm Productivity	mixed	Overall productivity and structural contradictions in production relations	Reading fidelity high Study strength speculative	0.02
In the production stage, workers lose decision-making power. Worker Satisfaction	negative	Workers' decision-making power	Reading fidelity high Study strength speculative	0.02
In the production stage, workers are disciplined by algorithms. Organizational Efficiency	negative	Algorithmic control/discipline over workers	Reading fidelity high Study strength speculative	0.02
In the production stage, workers are alienated into becoming data producers. Automation Exposure	negative	Role shift of workers toward producing data as labor	Reading fidelity high Study strength speculative	0.02
In the distribution phase, behavioral data unconsciously generated by workers drives algorithmic iteration yet remains excluded from the distribution system, resulting in hidden data exploitation. Labor Share	negative	Value distribution of data contributions (hidden data exploitation)	Reading fidelity high Study strength speculative	0.02
In the exchange phase, high barriers to entry for technology and capital foster market monopolies. Market Structure	negative	Market concentration / monopoly formation	Reading fidelity high Study strength low	0.06
In the consumption phase, high costs lead to service stratification, making it difficult for technological dividends to benefit the general public. Consumer Welfare	negative	Distribution of benefits / access to services (service stratification, consumer access)	Reading fidelity high Study strength speculative	0.02
The root causes of these problems include the disruption of labor relations boundaries by the transformation of the means of production, the exclusion of implicit data labor from distribution rules, the concentration of capital driven by high industry barriers, and social structural constraints on technological dissemination. Governance And Regulation	negative	Structural causes of inequality and power concentration in human-machine collaboration	Reading fidelity high Study strength speculative	0.02
To address these dilemmas, coordinated reconstruction of production relations is needed across three levels: macro-level institutional constraints, meso-level organizational transformation, and micro-level rights protection (e.g., recognition of data labor rights, anti-monopoly regulation, and algorithmic transparency). Governance And Regulation	positive	Policy and institutional change to improve equity and justice in human-machine collaboration	Reading fidelity high Study strength speculative	0.02