Labor Economics Of Autonomous Agent Integration - Displacement And Creation

Author: LawSphereX

Category: Labour & Employment

Published: 2026-08-23T20:28:15.255+00:00

Updated: 2026-08-30T07:09:22.024Z

LABOR ECONOMICS OF AUTONOMOUS AGENT

INTEGRATION: DISPLACEMENT AND CREATION

ABSTRACT

The integration of autonomous agents into the global workforce marks a pivotal ontological shift from reactive “Generative AI” to proactive, goaloriented “Agentic AI,” fundamentally altering the labor-capital compact. This article provides an exhaustive analysis of the labor economics driving this transition, contrasting the diminishing “reinstatement effect” of new task creation against the accelerating “displacement effect” of algorithmic substitution. We investigate the “Turing Trap,” where excessive investment in human-mimicking automation leads to the “unbossing” of middle management and the hollowing out of cognitive professions, potentially stagnating total factor productivity (TFP). Simultaneously, we explore the evolution of employment law from the 20th-century Industrial Relations model to the nascent regime of “Algorithmic Management.

” We analyze how the “fissured workplace” allows firms to use agents to exert direct control over workers while evading legal liability, challenging doctrines of respondeat superior and fiduciary duty. By synthesizing global case studiesfrom “digital colonialism” in the Global South to “technowelfare” in Japanthis article proposes a new regulatory framework centered on “Human-in- Command” principles and “duty-bearing” legal status for AI, ensuring that the rise of autonomous agents serves to reinstate, rather than obsolete, the value of human labor. Keywords: Autonomous Agents, Algorithmic Management, Reinstatement Effect, Respondeat Superior, Digital Taylorism, Turing Trap 1 Assistant Professor, School of Legal Studies, Babu Banarasi Das University, Lucknow, Uttar Pradesh, India.

I. Introduction: The Ontological Shift from Tool to Agent

The global economy stands at the precipice of a transformation that fundamentally diverges from previous iterations of technological change. While the last decade of digitization was defined by “Generative AI”- technologies that create content in reaction to human prompts- the current trajectory is rapidly shifting toward “Agentic AI. ” This distinction is not merely semantic but represents a profound rupture in the labor-capital compact. Generative AI acts as a tool, a passive instrument requiring a human operator to initiate and guide its function. In contrast, agentic AI possesses the capacity to proactively set goals, execute multi-step workflows, reason through ambiguity, and utilize external tools without constant human oversight (Korinek, 2024; McKinsey & Company, 2024).

As these systems evolve from “copilots” that assist to “autonomous agents” that execute, they challenge the foundational axioms of labor economics and the doctrinal pillars of employment law. We are witnessing the migration of automation from the factory floor to the manager’s office. Historically, automation substituted for labor in routine tasks while complementing it in non-routine cognitive tasks. However, the emergence of agents capable of planning, reasoning, and “unbossing” organizational structures suggests that the “reinstatement effect”- the economic mechanism by which technology creates new, higher-value tasks for humans- may be faltering (Acemoglu & Restrepo, 2019).

We face a potential “Turing Trap,” where investment is channeled into human-mimicking automation that suppresses labor’s bargaining power, rather than augmentation technologies that expand human capability (Brynjolfsson, 2022). Simultaneously, the legal frameworks governing work, forged in the industrial fires of the 20th century, are straining under the weight of “algorithmic management. ” The traditional employer-employee relationship, mediated by human supervisors and collective bargaining, is being supplanted by opaque, data-driven control systems. These systems facilitate a “fissured workplace” where liability is evaded through subcontracting, yet control is exerted with digital precision (Weil, 2014).

From the “digital Taylorism” of warehouse surveillance to the algorithmic discrimination in white-collar hiring, the law struggles to locate accountability when the “manager” is a non-human agent. This paper provides an exhaustive analysis of these dynamics, synthesizing economic theory, legal jurisprudence, and global case studies to map the future of work in the age of the autonomous agent.

II. The Macroeconomics of Autonomous Integration

2.1 The Task-Based Model and the Reinstatement Deficit

To understand the economic implications of autonomous agents, one must ground the analysis in the task-based framework developed by Acemoglu and Restrepo. This model posits that the production of goods and services consists of a bundle of tasks. Technological progress affects labor demand through three primary channels: the displacement effect, where capital replaces labor in specific tasks; the productivity effect, where lower costs drive expansion and labor demand in non-automated tasks; and the reinstatement effect, where technology creates entirely new tasks in which labor holds a comparative advantage (Acemoglu & Restrepo, 2019; Autor, 2024). Historically, the reinstatement effect has counterbalanced displacement. During the Industrial Revolution, while mechanization displaced weavers, it created new roles for engineers, machinists, and managers.

However, recent empirical evidence suggests a decoupling. Acemoglu (2024) indicates that the reinstatement effect has weakened significantly over the last three decades. Automation has accelerated, but the creation of new tasks that require human judgment and cannot be easily automated has lagged. The integration of autonomous agents exacerbates this “reinstatement deficit. ” Unlike previous waves of automation that targeted routine manual or clerical labor, agentic AI targets nonroutine cognitive tasks- planning, coding, legal research, and middle management- previously considered safe havens for human labor (Eloundou et al. , 2023). The “so-so automation” problem arises when technologies are good enough to displace workers but not productive enough to generate the massive surplus value required to reabsorb displaced labor into new sectors (Acemoglu & Restrepo, 2019).

2.2 The “Turing Trap” and the Productivity Paradox

A critical insight from the literature is the concept of the “Turing Trap,” articulated by Erik Brynjolfsson (2022). This theory suggests that the current trajectory of AI research and investment is overly obsessed with “human-like” intelligence- passing the Turing Test- rather than “machine-like” intelligence that augments human capabilities. When AI is designed to mimic humans (e. g. , a humanoid robot or an agent that writes code exactly like a human), it becomes a direct substitute for labor, depressing wages and shifting power from workers to capital owners. If the primary economic function of autonomous agents is to replace human labor with a cheaper digital equivalent, the result is a transfer of wealth from wages to capital rents, without necessarily triggering the massive productivity explosions seen in the age of electrification or the combustion engine.

Acemoglu’s (2024) projections are sobering in this regard: he estimates that the total factor productivity (TFP) gains from AI over the next decade may be “nontrivial but modest,” potentially less than 0. 66%. This contradicts the “singularity” narratives often promoted by technology optimists. If productivity gains are modest, they will be insufficient to drive the labor demand needed to offset the rapid displacement caused by agents capable of autonomous execution.

2.3 The “Unbossing” of the Corporation

Perhaps the most distinct labor market phenomenon identified in the 2024-2025 literature is the “unbossing” trend- the systematic elimination of middle management layers through the deployment of agentic AI (Gartner, 2024). Traditionally, middle managers served as the information routers and coordinators of the firm, assigning tasks, monitoring performance, and reporting to executive leadership. Autonomous agents, which can communicate with each other, track workflow states, and optimize resource allocation in real-time, render many of these functions obsolete (Tallam, 2025). Recent reports indicate that up to 50% of middle management roles could be displaced or fundamentally altered by 2026 (Gartner, 2024). Companies are moving toward “flat” structures where a small number of senior “orchestrators” direct vast fleets of AI agents and a dispersed layer of frontline workers.

This “hollowing out” of the corporate ladder disrupts the traditional pathway for career progression. If the entry-level tasks are automated (the “learning” roles) and the supervisory tasks are automated (the “managing” roles), the mechanism for human capital development breaks down, creating a long-term skills gap (Padisetti, 2024). III. The “Reinstatement” Reality: New Tasks and Skills

3.1 Emergent Roles

Orchestrators and Composers While the displacement effects are visible and statistically significant, the literature also documents the emergence of new, high-value tasks. The transition to “Agentic Organizations” is creating demand for roles that did not exist five years ago. Central among these is the “Agent Orchestrator” (McKinsey & Company, 2024; Unily, 2025). Unlike a prompt engineer who interacts with a model linearly, an Orchestrator designs and manages ecosystems of interacting agents. They function like a conductor, assigning specific roles (e. g. , “researcher,” “critic,” “coder”) to different agents, defining the protocols for their collaboration, and adjudicating conflicts in their outputs.

Cognate roles identified in the literature include: • Software Composer: A shift from “coding” to “composing,” where the human developer integrates modules generated by agents, focusing on system architecture and logic rather than syntax (Faia Corp, 2025). • AI Ethics Compliance Officer: A role necessitated by the regulatory risks of autonomous systems, focusing on auditing agent decisions for bias, hallucination, and regulatory adherence (CareerLoop, 2025). • Human-in-the-Loop (HITL) Specialist: Workers who intervene when agents reach a confidence threshold below a certain limit, particularly in high-stakes domains like healthcare and finance (Mosqueira-Rey et al. , 2025).

3.2 The Shift in Competencies

From Information to Interpersonal The integration of agents is driving a profound shift in the valuation of human skills. The Stanford “Future of Work” analysis and World Bank reports suggest a declining premium on information-processing skills- analysis, data retrieval, and synthesis- as these are the exact capabilities where agents excel (World Bank, 2025; Shao et al., 2025). Conversely, there is a rising premium on interpersonal and organizational skills. In an agent-saturated environment, the uniquely human ability to negotiate, empathize, motivate, and navigate complex political or social dynamics becomes the scarce resource. The “Human Agency Scale” proposed in recent research suggests that while “H1” (fully autonomous) tasks are delegated to agents, “H3-H5” tasks requiring meaningful coordination and communication remain the domain of humans (Shao et al., 2024).

Table 1: Economic Differentiation between Generative and Agentic AI Systems Feature Generative AI (2022-2024) Agentic AI (2025-Present) Primary Content Generation (Text, Autonomous Goal Execution & Function Code, Image) Decision Making Labor Human-Initiated (Prompt $\to$ Agent-Initiated (Goal $\to$ Plan Interaction Response) $\to$ Action) Junior creatives, coders, Middle managers, project Targeted Roles support staff coordinators, analysts Economic Productivity Augmentation Structural Substitution (Labor- Effect (Capital-Deepening) Saving) Management “The Invisible Colleague” / Tool for the worker Style Manager of the worker Vendor/Agent liability Liability Model User responsibility (emerging) economy (e. g. , Uber, Deliveroo), it has aggressively permeated traditional sectors, facilitating a resurgence of Taylorism- often termed “Digital Taylorism” (O’Neil, 2016; Mignucci, 2025).

In this regime, work is atomized into micro-tasks, and workers are subjected to continuous, granular surveillance. Wearable sensors in warehouses, keystroke loggers for remote workers, and biometric monitoring in logistics allow agents to optimize labor intensity in real-time. This creates a “panoptic” workplace where the asymmetry of information and power is extreme. Unlike human managers, algorithmic agents do not negotiate; they optimize. This rigidity creates severe friction with established labor standards that assume a degree of human discretion and dialogue (Rogers, 2022).

4.2 The Fissured Workplace and Liability Shields

Autonomous agents accelerate the trend David Weil (2014) identified as the “Fissured Workplace. ” Large lead firms shed direct employment relationships, relying on subcontracting, franchising, and outsourcing to reduce costs and liability. Agents are the technological glue that makes this possible. A lead firm can use an AI system to dictate precise productivity standards, delivery routes, and pricing to a subcontractor’s workforce without technically “employing” them (Weil, 2014). This structure allows firms to exercise control without responsibility. For example, an algorithm might set a delivery schedule that necessitates speeding, but if an accident occurs, the lead firm claims the driver is an independent contractor or the employee of a staffing agency.

The Department of Labor (DOL) and National Labor Relations Board (NLRB) are increasingly scrutinizing these arrangements, arguing that algorithmic control constitutes “joint employment,” but the legal tests are often outpaced by the technological reality (NLRB, 2022).

4.3 The Crisis of Collective Bargaining

The rise of AM poses an existential threat to collective bargaining. Agents can be programmed to detect and deter organizing activity. Legal filings against Amazon have alleged the use of “intrusive algorithms” to monitor workforce sentiment and identify unionization threats (Sainato, 2024). Furthermore, the “unbossing” effect removes the human manager who might historically have been a point of negotiation or grievance resolution. When the “boss” is code, there is no one to bargain with.

Unions are responding by attempting to “negotiate the algorithm.” This involves demanding transparency regarding the parameters of AI management systems, ensuring data rights, and establishing “Human-in-Command” clauses that prevent fully automated disciplinary actions (De Stefano, 2019; UNI Global Union, 2024). V. The Legal Frontier: Liability, Personhood, and Duty

5.1 The Crisis of Respondeat Superior and Mobley v. Workday

A central pillar of employment law is the doctrine of respondeat superior (let the master answer), which holds employers vicariously liable for the actions of their employees performed within the scope of employment. Autonomous agents rupture this doctrine. If an AI hiring agent discriminates against older applicants, or an autonomous vehicle causes harm, is the deployer liable? The recent case of Mobley v. Workday (2024) represents a seismic shift in this landscape. The U. S. District Court for the Northern District of California ruled that Workday, a vendor of AIbased HR software, could potentially be held liable as an “agent” of the employers who use its tools. The court noted that “drawing an artificial distinction between software decisionmakers and human decisionmakers would potentially gut anti-discrimination laws in the modern era” (Mobley v. Workday, Inc. , 2024; EEOC, 2024).

This ruling suggests that software vendors can no longer claim to be neutral tool providers; if their agents act with discriminatory impact, they share liability with the employer. This expands the scope of Title VII, the ADA, and the ADEA to include the algorithmic supply chain.

5.2 The “Right to an Explanation” and Ola Cabs

If Workday is about liability, the litigation against Ola Cabs in the Netherlands was about transparency. Drivers challenged the platform’s use of algorithms to deduct earnings and terminate accounts based on “fraud probability scores. ” The Amsterdam Court of Appeal ruled in favor of the drivers, finding that these “robo-firing” decisions constituted “automated decision-making” under Article 22 of the GDPR (Amsterdam Court of Appeal, 2023). Crucially, the court rejected Ola’s argument that revealing the algorithm’s logic would compromise trade secrets. It ruled that the workers’ right to transparency regarding their livelihood- specifically the ability to understand why they were fired- outweighed the company’s intellectual property interests.

This establishes a “right to understand” the algorithmic manager, mandating a level of explainability that many “black box” deep learning models currently fail to provide.

5.3 The Regulatory Response

The EU AI Act The European Union has taken the lead in regulating these risks with the AI Act. The Act explicitly categorizes AI systems used in employment, worker management, and access to selfemployment as “High Risk” (EU AI Act, Annex III, 2024). This classification imposes stringent obligations: • Transparency: Deployers must understand how the system makes decisions. • Human Oversight: The system must allow for human intervention (“Human-in-the- Loop”). • Data Governance: Training data must be vetted for bias (Clifford Chance, 2024). Table 2: Regulatory Divergence in Algorithmic Management Jurisdiction Key Regulation/Case Core Principle Impact on Agents Jurisdiction Key Regulation/Case Core Principle Impact on Agents Transparency: “Robo-firing” requires explainability. Trade secrets cannot block worker access to “fraud scores” or logic.

Netherlands Ola Cabs (2023) Source: Compiled by author from legal filings and regulatory texts. VI. Global Divergences: Technowelfare vs. Digital Colonialism

6.1 Digital Colonialism and the Global South

The “autonomy” of agents in the Global North is often subsidized by the invisible labor of the Global South. This dynamic is increasingly described as “Digital Colonialism” (Kwet, 2019; POSADA, 2022). The training of Large Language Models (LLMs) and agentic systems requires vast amounts of Reinforcement Learning from Human Feedback (RLHF). This worktagging images, flagging toxicity, correcting code- is outsourced to workers in Kenya, the Philippines, and India, often for pennies and under psychologically traumatizing conditions. This structure creates a new international division of labor. The “metropole” (Silicon Valley, London) captures the high-value “orchestration” and IP rights, while the “periphery” provides the raw data extraction and cleaning labor required to maintain the illusion of AI autonomy.

The ILO and labor advocates argue that these workers, often managed algorithmically via platforms, lack basic labor protections, effectively functioning as a “ghost workforce” for the AI economy (ILO, 2025).

6.2 Japan

Technowelfare and Demographic Necessity In contrast to the displacement anxieties of the West, Japan offers a divergent case study driven by demography. Facing a shrinking workforce and a super-aging society, Japan views autonomous agents not as a threat to labor, but as a savior. The concept of “Technowelfare” frames the integration of robotics and AI agents in elderly care as an ethical and economic necessity (Wright, 2023).

Here, the “substitution effect” is the goal. Agents are deployed to fill gaps in nursing, logistics, and service that cannot be met by human labor. The focus is less on “unbossing” and more on “augmentation” of the remaining workforce to prevent systemic collapse. This suggests that the labor impact of autonomous agents is not universal but is heavily mediated by a nation’s demographic profile and labor supply elasticity.

6.3 India

The Services Disruption India stands at a crossroads. As the “back office of the world,” its IT services sector is highly exposed to the automation of coding and BPO (Business Process Outsourcing) tasks by agents. However, reports indicate that over 80% of Indian enterprises are actively exploring autonomous agents to move up the value chain (Deloitte, 2025; Mishra, 2025). The challenge for India is whether it can transition its massive workforce from “managed” roles (executing specs) to “orchestrator” roles (managing agents) before the window of opportunity closes. VII. Conclusion The rise of autonomous agents represents the “Algorithmic Leviathan”- a force that promises immense productivity but threatens to consume the structures of human agency and employment protection.

The economic evidence suggests we are currently in a “reinstatement deficit,” where the displacement of human tasks outpaces the creation of new ones, exacerbated by a tax system that favors silicon over sweat. The resulting “unbossing” of corporations and the spread of “digital Taylorism” create a precarious future for the workforce, bifurcated into a small elite of orchestrators and a mass of algorithmically managed subjects. Legal frameworks are scrambling to catch up. The ruling in Mobley v. Workday and the EU AI Act represent the first concrete steps toward taming this Leviathan, establishing that liability cannot be outsourced to an algorithm and that the data rights of workers are paramount. The path forward requires a dual approach: Economic policy must end the “automation subsidy” and invest heavily in “human-complementary” R&D to escape the Turing Trap.

Legal policy must enshrine “Human-in-Command” principles, ensuring that the “invisible colleague” remains a tool, not a tyrant. Ultimately, the integration of autonomous agents is not a deterministic outcome but a choice. We can design agents that act as “copilots,” augmenting human capacity and broadening the middle class, or we can design agents that act as “scabs,” undermining labor power and concentrating wealth. The difference will not be determined by the code, but by the laws and values we impose upon it.

References

Acemoglu, D. (2024). The simple macroeconomics of AI (NBER Working Paper No. 32487). National Bureau of Economic Research. https://doi. org/10. 3386/w32487 Acemoglu, D. , & Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives, 33(2), 3-30. https://doi. org/10. 1257/jep. 33. 2. 3 Acemoglu, D. , Manera, A. , & Restrepo, P. (2020). Does the US tax code favor automation? Brookings Papers on Economic Activity, 2020(1), 231-300. https://www. brookings. edu/articles/does-the-us-tax-code-favor-automation/ Amsterdam Court of Appeal. (2023, April 4). Uber/Ola Cabs v. App Drivers & Couriers Union (Case No. 200. 295. 743/01). Rechtspraak. nl. Autor, D. H. (2024). The labor market impacts of technological change: From unbridled enthusiasm to qualified optimism to vast uncertainty (NBER Working Paper No. 30074).

National Bureau of Economic Research. https://doi. org/10. 3386/w30074 Brynjolfsson, E. (2022). The Turing trap: The promise & peril of human-like artificial intelligence. Daedalus, 151(2), 272-287. https://doi. org/10. 1162/daed_a_01915 Brynjolfsson, E. , Li, D. , & Raymond, L. (2023). Generative AI at work (NBER Working Paper No. 31161). National Bureau of Economic Research. https://doi. org/10. 3386/w31161 CareerLoop. (2025). Future career paths: AI Ethics Compliance Officer. CareerLoop Insights. https://careerloop. app/careers Clifford Chance. (2024, August). What does the EU AI Act mean for employers? Clifford Chance Briefings. https://www. cliffordchance. com De Stefano, V. (2019). ‘Negotiating the algorithm’: Automation, artificial intelligence and labour protection. Comparative Labor Law & Policy Journal, 41(1), 15-46. Deloitte. (2025).

State of GenAI in India: From experimentation to execution. Deloitte Insights.

Duggan, J. , Sherman, U. , Carbery, R. , & McDonnell, A. (2020). Algorithmic management and app-work in the gig economy: A research agenda for employment relations and HRM. Human Resource Management Journal, 30(1), 114-132. https://doi. org/10. 1111/1748-8583. 12258 Eloundou, T. , Manning, S. , Mishkin, P. , & Rock, D. (2023). GPTs are GPTs: An early look at the labor market impact potential of large language models. arXiv. https://doi. org/10. 48550/arXiv. 2303. 10130 Equal Employment Opportunity Commission. (2024, April 9). Brief of the EEOC as Amicus Curiae in Support of Plaintiff-Appellant, Mobley v. Workday, Inc. (No. 3:23-cv-00770-RFL). United States District Court for the Northern District of California. European Parliament & Council of the European Union. (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act).

Official Journal of the European Union. Faia Corp. (2025). The rise of orchestrators: New job titles in the age of AI. Faia Corp Insights. https://www. faiacorp. com/insights/the-rise-of-orchestrators Gartner. (2024). Predicts 2024: The future of work and the unbossing of the enterprise. Gartner Research. Gusto. (2024). The rise of the unbossed economy: SMB employment trends report. Gusto Economic Research. International Labour Organization. (2025). Global case studies of social dialogue on AI and algorithmic management (Working Paper 144). ILO. https://doi. org/10. 54394/voqe4924 Korinek, A. (2024). AI agents for economic research. Journal of Economic Literature (forthcoming). https://www. aeaweb. org/articles? id=10. 1257/jel. 20231736 Kwet, M. (2019). Digital colonialism: US empire and the new imperialism in the Global South. Race & Class, 60(4), 3-26. https://doi. org/10.

1177/0306396818823172 McKinsey & Company. (2024). The agentic organization: Contours of the next paradigm for the AI era. McKinsey Global Institute.

Mignucci, M. (2025). Extraordinary times: Algorithmic management and labor law. Columbia Journal of Law and Social Problems, 58, 397. Mishra, S. (2025, May 31). AI’s great disruption in India’s BFSI: 15 ways automation will slash 90% of jobs. Medium. Mobley v. Workday, Inc., No. 3:23-cv-00770-RFL (N.D. Cal. July 12, 2024). Mosqueira-Rey, E., Hernandez-Pereira, E., Alonso-Betanzos, A., Alonso-Moral, G., & Flores- Vergara, I. (2025). Human-in-the-loop machine learning: A state of the art classification and review. Artificial Intelligence Review, 58, 1-54. National Labor Relations Board, Office of the General Counsel. (2022, October 31). Electronic monitoring and algorithmic management of employees interfering with the exercise of Section

7 rights (Memorandum GC 23-02).

O’Neil, C. (2016). Weapons of math destruction: How big data increases inequality and threatens democracy. Crown. Padisetti, M. (2024, November 27). Reimagining middle management in the era of AI. Forbes. Parent-Rocheleau, X., & Parker, S. K. (2022). Algorithms as work designers: How algorithmic management influences the quality of jobs. Human Resource Management Review, 32(3), 100838. Posada, J. (2022). The coloniality of data work. Media, Culture & Society, 44(8), 1439-1455. https://doi.org/10.1177/01634437221111935 Rogers, B. (2022). Workplace data and democracy. Georgetown Law Technology Review, 6, 454. Sainato, M. (2024, May 21). Amazon surveillance lawsuit union. The Guardian. Salesforce. (2025). Agentic AI vs Generative AI: Understanding the difference. Salesforce Agentforce.

Shao, Y. , Zhang, W. , & Wang, H. (2024). The human agency scale: Auditing AI integration in the workplace. arXiv. https://doi. org/10. 48550/arXiv. 2406. 06576 Tallam, K. (2025). From autonomous agents to integrated systems: Orchestrated distributed intelligence. arXiv. https://doi. org/10. 48550/arXiv. 2503. 13754 Uber BV v. Aslam, UKSC 5. UNI Global Union. (2024). Algorithmic management: Opportunities for collective action. UNI Global Union. Unily. (2025, September 9). Unily launches Agent Orchestrator to unlock the multi-agent future of work. Unily Newsroom. Weil, D. (2014). The fissured workplace: Why work became so bad for so many and what can be done to improve it. Harvard University Press. World Economic Forum. (2025). Jobs of tomorrow: Technology and the future of the world’s largest workforces. World Economic Forum. Wright, J. (2023).

Robots won’t save Japan: An ethnography of elder care automation. Cornell University Press.