ETR-2026-07
Accelerating the Dystopia
Why Artificial Intelligence Cannot Save a System We Refuse to Fix
Citations
The source registry for this essay. Source types, roles, and notes describe the evidence; they do not rank its quality.
- S101
Shrouded Attributes, Consumer Myopia, and Information Suppression in Competitive Markets
- Author or institution
- Xavier Gabaix; David Laibson
- Year
- 2005
- Source type
- peer-reviewed economics / NBER
- Authority role
- mechanism evidence
- Source families
- AF01;AF05
Shows profitable informational shrouding can persist in competitive markets and generate allocative inefficiency.
Open source: Shrouded Attributes, Consumer Myopia, and Information Suppression in Competitive Markets ↗ - S102
Bringing Dark Patterns to Light
- Author or institution
- Federal Trade Commission
- Year
- 2022
- Source type
- official regulator report
- Authority role
- documented practice
- Source families
- AF01;AF05
Documents obscured material information, buried fees, difficult cancellation, disguised ads, and privacy manipulation.
Open source: Bringing Dark Patterns to Light ↗ - S103
Alan Greenspan Oral History Interview
- Author or institution
- Federal Reserve Board
- Year
- 2009
- Source type
- official institutional oral history
- Authority role
- historical primary evidence
- Source families
- AF01;AF06
Greenspan describes a conscious effort toward constructive ambiguity; motivation included market functioning.
Open source: Alan Greenspan Oral History Interview ↗ - S104
Commensuration as a Social Process
- Author or institution
- Wendy Nelson Espeland; Mitchell L. Stevens
- Year
- 1998
- Source type
- peer-reviewed sociology
- Authority role
- conceptual mechanism
- Source families
- AF01;AF02
Explains comparison through common metrics as a cognitive and political transformation with power effects.
Open source: Commensuration as a Social Process ↗ - S105
Rankings and Reactivity: How Public Measures Recreate Social Worlds
- Author or institution
- Wendy Nelson Espeland; Michael Sauder
- Year
- 2007
- Source type
- peer-reviewed sociology
- Authority role
- mechanism evidence
- Source families
- AF02
Shows rankings can change organizations through commensuration and self-fulfilling processes.
Open source: Rankings and Reactivity: How Public Measures Recreate Social Worlds ↗ - S106
Magic Quadrants Research Methodology
- Author or institution
- Gartner
- Year
- 2026
- Source type
- vendor methodology
- Authority role
- first-party description
- Source families
- AF02
Defines Ability to Execute and Completeness of Vision axes; Gartner itself cautions MQ is a starting point, not a universal best-fit ranking.
Open source: Magic Quadrants Research Methodology ↗ - S107
Engineering Culture: Control and Commitment in a High-Tech Corporation
- Author or institution
- Gideon Kunda
- Year
- 1992
- Source type
- ethnographic monograph
- Authority role
- organizational mechanism
- Source families
- AF03
Classic ethnography of normative control targeting commitment, identity, and the employee's sense of self.
Open source: Engineering Culture: Control and Commitment in a High-Tech Corporation ↗ - S108
Tightening the Iron Cage: Concertive Control in Self-Managing Teams
- Author or institution
- James R. Barker
- Year
- 1993
- Source type
- peer-reviewed organizational research
- Authority role
- organizational mechanism
- Source families
- AF03
Shows value-based peer norms can become more constraining than prior bureaucratic control.
Open source: Tightening the Iron Cage: Concertive Control in Self-Managing Teams ↗ - S109
Why Employees Do Bad Things: Moral Disengagement and Unethical Organizational Behavior
- Author or institution
- Celia Moore et al.
- Year
- 2012
- Source type
- peer-reviewed organizational psychology
- Authority role
- behavioral evidence
- Source families
- AF03
Moral disengagement predicts multiple forms of unethical organizational behavior across studies.
Open source: Why Employees Do Bad Things: Moral Disengagement and Unethical Organizational Behavior ↗ - S110
Leaders' organizational identification, moral disengagement, and trust in followers engaging in unethical pro-organizational behavior
- Author or institution
- Sebastian C. Schuh et al.
- Year
- 2021
- Source type
- peer-reviewed HR research
- Authority role
- behavioral evidence
- Source families
- AF03
Connects organizational identification and moral disengagement with leader trust in unethical pro-organizational behavior.
Open source: Leaders' organizational identification, moral disengagement, and trust in followers engaging in unethical pro-organizational behavior ↗ - S111
Financialization and corporate downsizing as a shareholder value strategy
- Author or institution
- Jiwook Jung; Young-Mi Lee
- Year
- 2022
- Source type
- peer-reviewed sociology/economics
- Authority role
- institutional evidence
- Source families
- AF03;AF04
Finds large US firms more likely to downsize after missing analyst forecasts; institutional ownership and CFO presence amplify association.
Open source: Financialization and corporate downsizing as a shareholder value strategy ↗ - S112
How pervasive is corporate fraud?
- Author or institution
- Alexander Dyck; Adair Morse; Luigi Zingales
- Year
- 2024
- Source type
- peer-reviewed accounting/economics
- Authority role
- detection-bias evidence
- Source families
- AF03
Natural-experiment estimates imply roughly one-third of corporate fraud is detected and about 10% of large public firms commit securities fraud annually, with stated uncertainty.
Open source: How pervasive is corporate fraud? ↗ - S113
Who Blows the Whistle on Corporate Fraud?
- Author or institution
- Alexander Dyck; Adair Morse; Luigi Zingales
- Year
- 2007
- Source type
- NBER / peer-reviewed lineage
- Authority role
- detection mechanism
- Source families
- AF03
Shows detection is distributed across employees, media, regulators, auditors and other actors rather than one central monitor.
Open source: Who Blows the Whistle on Corporate Fraud? ↗ - S114
Automation and New Tasks: How Technology Displaces and Reinstates Labor
- Author or institution
- Daron Acemoglu; Pascual Restrepo
- Year
- 2019
- Source type
- NBER / peer-reviewed economics
- Authority role
- labor mechanism
- Source families
- AF04;AF14
Task framework: automation displaces labor and reduces labor share; new tasks can reinstate labor demand.
Open source: Automation and New Tasks: How Technology Displaces and Reinstates Labor ↗ - S115
Tasks, Automation, and the Rise in US Wage Inequality
- Author or institution
- Daron Acemoglu; Pascual Restrepo
- Year
- 2021
- Source type
- NBER / peer-reviewed economics
- Authority role
- empirical labor evidence
- Source families
- AF04
Authors estimate 50-70% of changes in US wage structure over four decades are accounted for by relative declines among routine-task groups in rapidly automating industries.
Open source: Tasks, Automation, and the Rise in US Wage Inequality ↗ - S116
Does the U.S. Tax Code Favor Automation?
- Author or institution
- Daron Acemoglu; Andrea Manera; Pascual Restrepo
- Year
- 2020
- Source type
- NBER economics
- Authority role
- policy mechanism
- Source families
- AF04;AF06
Models tax bias favoring capital as producing excessive automation relative to optimal-tax counterfactual.
Open source: Does the U.S. Tax Code Favor Automation? ↗ - S117
Productivity and Costs, Second Quarter 2026, Revised
- Author or institution
- U.S. Bureau of Labor Statistics
- Year
- 2026
- Source type
- official statistics
- Authority role
- current empirical baseline
- Source families
- AF04;AF14
Reports labor share of nonfarm business sector output at 52.8%, lowest in series beginning 1947; real hourly compensation down 0.1% over four quarters.
Open source: Productivity and Costs, Second Quarter 2026, Revised ↗ - S118
Distributional Financial Accounts
- Author or institution
- Federal Reserve Board
- Year
- 2026
- Source type
- official statistics
- Authority role
- wealth-distribution baseline
- Source families
- AF04;AF06;AF14
Q1 2026 data imply top 10% hold roughly 68% of household net worth and bottom 50% roughly 2.5%, based on category totals.
Open source: Distributional Financial Accounts ↗ - S119
Income in the United States: 2025
- Author or institution
- U.S. Census Bureau
- Year
- 2026
- Source type
- official statistics
- Authority role
- credibility check / baseline
- Source families
- AF04;AF14
Median household income rose in 2025; useful guard against simplistic claims that all households are continuously getting poorer.
Open source: Income in the United States: 2025 ↗ - S121
Incumbency Advantages and Consumer Search Costs
- Author or institution
- Klaus Gugler et al.
- Year
- 2023
- Source type
- peer-reviewed industrial organization
- Authority role
- consumer-inertia mechanism
- Source families
- AF05
Evidence that incumbents can price discriminate around consumer search costs, with nonsearching/high-search-cost consumers facing higher baseline tariffs.
Open source: Incumbency Advantages and Consumer Search Costs ↗ - S122
Tackling the loyalty penalty
- Author or institution
- UK Competition and Markets Authority
- Year
- 2018
- Source type
- official competition regulator
- Authority role
- documented practice
- Source families
- AF05
Documents longstanding customers paying materially more than new customers in multiple markets due to price jumps, inertia, confusion, and switching frictions.
Open source: Tackling the loyalty penalty ↗ - S123
Personalized Pricing and Customer Welfare
- Author or institution
- Jean-Pierre Dubé; Sanjog Misra
- Year
- 2023
- Source type
- peer-reviewed economics
- Authority role
- personalized-pricing evidence
- Source families
- AF05
Field setting found ML personalized pricing raised expected profits relative to optimized uniform pricing and reduced aggregate consumer surplus, while benefits varied across consumers.
Open source: Personalized Pricing and Customer Welfare ↗ - S124
FTC Surveillance Pricing Study Indicates Wide Range of Personal Data Used to Set Individualized Consumer Prices
- Author or institution
- Federal Trade Commission
- Year
- 2025
- Source type
- official regulator study
- Authority role
- current practice evidence
- Source families
- AF05
FTC found intermediaries using location, browsing, shopping behavior, mouse movements, carts, demographics and other data to tailor prices, discounts, and product exposure for at least 250 clients.
Open source: FTC Surveillance Pricing Study Indicates Wide Range of Personal Data Used to Set Individualized Consumer Prices ↗ - S125
Shrinkflation: Consumer Search and Firm Pricing
- Author or institution
- Marketing Science
- Year
- 2025
- Source type
- peer-reviewed marketing/economics
- Authority role
- shrinkflation evidence
- Source families
- AF05
Finds downsizing more common than upsizing and consumers more responsive to price than package size, creating incentive to raise effective unit price through size changes.
Open source: Shrinkflation: Consumer Search and Firm Pricing ↗ - S126
The case of shrinking packaging
- Author or institution
- Federal Trade Commission
- Year
- 2024
- Source type
- official consumer guidance
- Authority role
- practice description
- Source families
- AF05
Explains how familiar packaging can make quantity reductions less salient to consumers.
Open source: The case of shrinking packaging ↗ - S127
Broadband Consumer Labels
- Author or institution
- Federal Communications Commission
- Year
- 2024
- Source type
- official regulation
- Authority role
- transparency intervention
- Source families
- AF05
Requires broadband labels that include introductory and post-introductory pricing and other standardized information.
Open source: Broadband Consumer Labels ↗ - S128
Treasury Announces Increased Sizes of Nominal Long-End Liquidity Support Buybacks Beginning September 9
- Author or institution
- U.S. Treasury
- Year
- 2026
- Source type
- official policy announcement
- Authority role
- current primary source
- Source families
- AF06
Long-end liquidity-support buybacks increased from $2B maximum per operation to at least $4B through the quarter.
Open source: Treasury Announces Increased Sizes of Nominal Long-End Liquidity Support Buybacks Beginning September 9 ↗ - S129
Quarterly Refunding Statement, August 5 2026
- Author or institution
- U.S. Treasury
- Year
- 2026
- Source type
- official policy statement
- Authority role
- current primary source
- Source families
- AF06
Maintains coupon/FRN auction sizes; marginal financing uses bills/CMBs and other auctions; projects up to $38B off-the-run liquidity buybacks plus up to $25B short-maturity cash-management buybacks.
Open source: Quarterly Refunding Statement, August 5 2026 ↗ - S130
Treasury Supply Shocks: Propagation Through Debt Expansion and Maturity Adjustment
- Author or institution
- Huixin Bi; Maxime Phillot; Sarah Zubairy / Federal Reserve Bank of Kansas City
- Year
- 2026
- Source type
- Fed research working paper
- Authority role
- mechanism evidence
- Source families
- AF06
Finds Treasury maturity structure affects financial conditions and can reinforce or offset Fed asset purchases.
Open source: Treasury Supply Shocks: Propagation Through Debt Expansion and Maturity Adjustment ↗ - S131
Government Debt Management at the Zero Lower Bound
- Author or institution
- Robin Greenwood; Samuel Hanson; Joshua Rudolph; Lawrence Summers
- Year
- 2014
- Source type
- Treasury-hosted academic paper
- Authority role
- mechanism precedent
- Source families
- AF06
Argues Treasury maturity extension acted as reverse QE by putting duration back into private hands while Fed removed it.
Open source: Government Debt Management at the Zero Lower Bound ↗ - S132
The Mechanics of LSAP Transmission
- Author or institution
- Federal Reserve Bank of New York
- Year
- 2010
- Source type
- central-bank research/speech
- Authority role
- QE mechanism
- Source families
- AF06
Describes duration removal, term-premium, and portfolio-balance effects of large-scale asset purchases.
Open source: The Mechanics of LSAP Transmission ↗ - S133
The Federal Reserve's Balance Sheet and Treasury Markets
- Author or institution
- Michael S. Barr / Federal Reserve Board
- Year
- 2026
- Source type
- official central-bank speech
- Authority role
- QE mechanism/current framing
- Source families
- AF06
States duration-risk removal was central to QE's transmission at the effective lower bound.
Open source: The Federal Reserve's Balance Sheet and Treasury Markets ↗ - S134
The Case for Removing Mortgage-Backed Securities from the Federal Reserve's Balance Sheet
- Author or institution
- Christopher J. Waller / Federal Reserve Board
- Year
- 2025
- Source type
- official central-bank speech
- Authority role
- maturity-transform mechanism
- Source families
- AF06
Describes QE as financing longer-duration assets with short-term reserve liabilities, creating consolidated maturity exposure.
Open source: The Case for Removing Mortgage-Backed Securities from the Federal Reserve's Balance Sheet ↗ - S135
Unconventional Monetary Policies and Inequality
- Author or institution
- Federal Reserve Bank of New York
- Year
- 2024
- Source type
- central-bank staff report
- Authority role
- distributional mechanism
- Source families
- AF06
Model finds QE/forward guidance supported activity and reduced inequality within bottom 90 through employment, while widening top-vs-rest income gap through profits/equities.
Open source: Unconventional Monetary Policies and Inequality ↗ - S136
The Budget and Economic Outlook: 2026 to 2036
- Author or institution
- Congressional Budget Office
- Year
- 2026
- Source type
- official fiscal projection
- Authority role
- current fiscal baseline
- Source families
- AF06;AF14
Net interest rises from $1.0T/3.3% GDP in 2026 to $2.1T/4.6% GDP in 2036; public debt rises from 101% to 120% GDP.
Open source: The Budget and Economic Outlook: 2026 to 2036 ↗ - S137
COFER Data Brief, First Quarter 2026
- Author or institution
- International Monetary Fund
- Year
- 2026
- Source type
- official reserve statistics
- Authority role
- current reserve-currency baseline
- Source families
- AF06
Dollar reserve share increased from 56.42% in 2025Q4 to 57.13% in 2026Q1; IMF says share broadly stable, while gold's rise largely reflects valuation effects.
Open source: COFER Data Brief, First Quarter 2026 ↗ - S138
Contributions to Super PACs and Hybrid PACs
- Author or institution
- Federal Election Commission
- Year
- 2026
- Source type
- official election-law guidance
- Authority role
- current political-finance fact
- Source families
- AF07
Independent expenditure-only committees may accept unlimited contributions from individuals, corporations, labor organizations and other political committees, subject to source restrictions and reporting.
Open source: Contributions to Super PACs and Hybrid PACs ↗ - S139
Contributors' identities not subject to disclosure
- Author or institution
- Internal Revenue Service
- Year
- 2026
- Source type
- official tax-law guidance
- Authority role
- current disclosure fact
- Source families
- AF07
Most tax-exempt organizations generally need not publicly disclose contributor names/addresses on annual returns, with important exceptions including private foundations and certain 527 political organizations.
Open source: Contributors' identities not subject to disclosure ↗ - S140
2025 Lobbying Disclosure: Observations on Compliance with Requirements
- Author or institution
- U.S. Government Accountability Office
- Year
- 2026
- Source type
- official audit/report
- Authority role
- current lobbying-compliance evidence
- Source families
- AF07
Most sampled lobbyists showed evidence of compliance, but 22% of quarterly reports did not fully disclose relevant prior federal jobs; Congress referred 12,391 filing cases to DOJ during 2016-2025.
Open source: 2025 Lobbying Disclosure: Observations on Compliance with Requirements ↗ - S141
Is It Whom You Know or What You Know?
- Author or institution
- Marianne Bertrand; Matilde Bombardini; Francesco Trebbi
- Year
- 2014
- Source type
- peer-reviewed political economy
- Authority role
- lobbying mechanism
- Source families
- AF07
Connections are central to lobbying; lobbyists follow connected politicians across committee changes and connection premiums exceed expertise premiums in the data.
Open source: Is It Whom You Know or What You Know? ↗ - S142
Importing Political Polarization? The Electoral Consequences of Rising Trade Exposure
- Author or institution
- David Autor; David Dorn; Gordon Hanson; Kaveh Majlesi
- Year
- 2020
- Source type
- peer-reviewed political economy
- Authority role
- economic-grievance evidence
- Source families
- AF08
Trade-shocked US labor markets experienced political realignment and reduced moderateness, with heterogeneous direction depending on local composition.
Open source: Importing Political Polarization? The Electoral Consequences of Rising Trade Exposure ↗ - S143
Like-minded sources on Facebook are prevalent but not polarizing
- Author or institution
- Brendan Nyhan et al.
- Year
- 2023
- Source type
- Nature field experiment
- Authority role
- counterevidence / platform mechanism
- Source families
- AF08
Reducing like-minded-source exposure by about one-third changed content exposure and reduced uncivil/misinformation-source exposure but produced no measurable change in eight preregistered political-attitude outcomes.
Open source: Like-minded sources on Facebook are prevalent but not polarizing ↗ - S144
Loneliness and Problematic Media Use: Meta-Analysis of Longitudinal Studies
- Author or institution
- Jia Yuin Fam; Niko Mannikko
- Year
- 2025
- Source type
- peer-reviewed longitudinal meta-analysis
- Authority role
- psychosocial evidence
- Source families
- AF08
26 studies / 24,798 people show small bidirectional longitudinal associations when using beta estimates, about 0.09-0.10.
Open source: Loneliness and Problematic Media Use: Meta-Analysis of Longitudinal Studies ↗ - S145
Doomscrolling evokes existential anxiety and fosters pessimism about human nature? Evidence from Iran and the United States
- Author or institution
- Reza Shabahang et al.
- Year
- 2024
- Source type
- peer-reviewed behavioral study
- Authority role
- psychosocial evidence
- Source families
- AF08
Cross-sectional convenience samples of Iranian (n=620) and U.S. (n=180) university-student social-media users; doomscrolling associated with existential anxiety in both samples. Associational evidence, not a population-wide causal estimate.
Open source: Doomscrolling evokes existential anxiety and fosters pessimism about human nature? Evidence from Iran and the United States ↗ - S146
Investigating affective use and emotional well-being on ChatGPT
- Author or institution
- OpenAI
- Year
- 2025
- Source type
- first-party large-scale usage study
- Authority role
- emerging evidence
- Source families
- AF08;AF09
Emotional use rare overall, concentrated among a small heavy-user subset; extended/high-intensity use and attachment associated with worse outcomes.
Open source: Investigating affective use and emotional well-being on ChatGPT ↗ - S147
How AI and Human Behaviors Shape Psychosocial Effects of Chatbot Use
- Author or institution
- MIT Media Lab
- Year
- 2025
- Source type
- longitudinal controlled study
- Authority role
- emerging causal/observational evidence
- Source families
- AF08
RCT did not find strong assigned-condition effects; heavier voluntary use correlated with worse psychosocial outcomes.
Open source: How AI and Human Behaviors Shape Psychosocial Effects of Chatbot Use ↗ - S148
Emotion shapes the diffusion of moralized content in social networks
- Author or institution
- William J. Brady et al.
- Year
- 2017
- Source type
- PNAS empirical study
- Authority role
- attention/diffusion mechanism
- Source families
- AF08
Moral-emotional words associated with increased diffusion, particularly within ideological group boundaries.
Open source: Emotion shapes the diffusion of moralized content in social networks ↗ - S149
How social learning amplifies moral outrage expression in online social networks
- Author or institution
- William J. Brady et al.
- Year
- 2021
- Source type
- Science Advances preregistered studies
- Authority role
- platform-learning mechanism
- Source families
- AF08
Positive social feedback predicts future outrage expression and users conform to network norms; combines observational and experimental evidence.
Open source: How social learning amplifies moral outrage expression in online social networks ↗ - S150
Estimating the effect size of moral contagion in online networks: a preregistered replication and meta-analysis
- Author or institution
- William J. Brady et al.
- Year
- 2025
- Source type
- peer-reviewed replication/meta-analysis / PNAS Nexus
- Authority role
- replication evidence
- Source families
- AF08
Pre-registered replication plus meta-analysis of 27 studies (N=4,821,006); reports IRR 1.13, i.e. 13% greater expected shares per additional moral-emotional word, with heterogeneity across studies/datasets.
Open source: Estimating the effect size of moral contagion in online networks: a preregistered replication and meta-analysis ↗ - S151
FTC Staff Report on AI Partnerships and Investments
- Author or institution
- Federal Trade Commission
- Year
- 2025
- Source type
- official competition study
- Authority role
- AI concentration evidence
- Source families
- AF09
Documents equity, revenue-sharing, cloud-spend commitments, consultation/control rights, exclusivities, information sharing, and switching/input-access concerns in major AI-cloud partnerships.
Open source: FTC Staff Report on AI Partnerships and Investments ↗ - S152
AI Index Report 2026: Research and Development
- Author or institution
- Stanford HAI
- Year
- 2026
- Source type
- academic annual report
- Authority role
- AI capability/concentration baseline
- Source families
- AF09;AF13
Reports industry produced more than 90% of notable AI models in 2025 and tracks rapid growth in AI compute; also notes transparency gaps among leading models.
Open source: AI Index Report 2026: Research and Development ↗ - S153
AI Index Report 2026: Economy
- Author or institution
- Stanford HAI
- Year
- 2026
- Source type
- academic annual report
- Authority role
- AI diffusion baseline
- Source families
- AF09;AF14
Survey evidence shows broad organizational AI adoption while agent deployment remains much narrower across functions.
Open source: AI Index Report 2026: Economy ↗ - S154
Competition in artificial intelligence infrastructure
- Author or institution
- OECD
- Year
- 2025
- Source type
- intergovernmental competition analysis
- Authority role
- market concentration evidence
- Source families
- AF09
Reviews concentration and barriers across chips, cloud and AI infrastructure; recent estimates place Nvidia above 80% in AI GPUs.
Open source: Competition in artificial intelligence infrastructure ↗ - S155
The Rapid Adoption of Generative AI
- Author or institution
- Alexander Bick; Adam Blandin; David Deming
- Year
- 2024
- Source type
- NBER / survey evidence
- Authority role
- adoption baseline
- Source families
- AF09;AF14
By late 2024, nearly 40% of US adults 18-64 had used generative AI and about 23% of employed respondents used it for work in the previous week.
Open source: The Rapid Adoption of Generative AI ↗ - S156
Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification
- Author or institution
- Joy Buolamwini; Timnit Gebru
- Year
- 2018
- Source type
- peer-reviewed ML audit
- Authority role
- historical bias evidence
- Source families
- AF10
Commercial classifiers showed stark intersectional performance disparities; benchmark composition itself was imbalanced.
Open source: Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification ↗ - S157
Semantics derived automatically from language corpora contain human-like biases
- Author or institution
- Aylin Caliskan; Joanna Bryson; Arvind Narayanan
- Year
- 2017
- Source type
- Science empirical study
- Authority role
- representation-bias evidence
- Source families
- AF10
Distributional semantic representations trained on ordinary language reproduce human-like historical associations.
Open source: Semantics derived automatically from language corpora contain human-like biases ↗ - S158
Towards a Standard for Identifying and Managing Bias in Artificial Intelligence
- Author or institution
- NIST SP 1270
- Year
- 2022
- Source type
- official technical report
- Authority role
- bias taxonomy/governance
- Source families
- AF10
Treats AI bias as socio-technical and distinguishes systemic, statistical/computational, and human biases.
Open source: Towards a Standard for Identifying and Managing Bias in Artificial Intelligence ↗ - S159
Low Frequency Names Exhibit Bias and Overfitting in Contextualizing Language Models
- Author or institution
- Wolfe; Caliskan
- Year
- 2021
- Source type
- EMNLP empirical study
- Authority role
- representation-bias evidence
- Source families
- AF10
Minority names are less frequent in training corpora and lower frequency is associated with unfavorable BERT associations in studied settings.
Open source: Low Frequency Names Exhibit Bias and Overfitting in Contextualizing Language Models ↗ - S160
The Token Tax: Systematic Bias in Multilingual Tokenization
- Author or institution
- Lundin et al.
- Year
- 2026
- Source type
- AfricaNLP empirical study
- Authority role
- tokenization-performance evidence
- Source families
- AF10
Across 10 LLMs and 16 African languages, higher token fertility consistently predicts lower accuracy; reasoning narrows but does not eliminate gap.
Open source: The Token Tax: Systematic Bias in Multilingual Tokenization ↗ - S161
Tokenization Falling Short: On Subword Robustness in Large Language Models
- Author or institution
- Chai et al.
- Year
- 2024
- Source type
- EMNLP empirical study
- Authority role
- tokenization-robustness evidence
- Source families
- AF10
Subword tokenization remains sensitive to perturbations such as typos and formatting; scaling mitigates but does not eliminate.
Open source: Tokenization Falling Short: On Subword Robustness in Large Language Models ↗ - S162
Understanding and Mitigating Tokenization Bias
- Author or institution
- Phan et al.
- Year
- 2024
- Source type
- research preprint
- Authority role
- tokenization-bias mechanism
- Source families
- AF10
Argues common tokenization procedures induce sampling biases that are not automatically eliminated by more training data.
Open source: Understanding and Mitigating Tokenization Bias ↗ - S164
Translating Neuralese
- Author or institution
- Jacob Andreas; Anca Dragan; Dan Klein
- Year
- 2017
- Source type
- ACL peer-reviewed paper
- Authority role
- emergent-communication precedent
- Source families
- AF11
Treats effective learned multi-agent communication as an interpretability problem and proposes translation into human language.
Open source: Translating Neuralese ↗ - S165
Language Grounded Multi-agent Reinforcement Learning with Human-interpretable Communication
- Author or institution
- Huao Li et al.
- Year
- 2024
- Source type
- NeurIPS / arXiv research paper
- Authority role
- emergent-communication evidence
- Source families
- AF11
Learned MARL communication is often not interpretable to humans or agents not co-trained together; language grounding improves interoperability while retaining performance in the reported setting.
Open source: Language Grounded Multi-agent Reinforcement Learning with Human-interpretable Communication ↗ - S166
Enabling Agents to Communicate Entirely in Latent Space
- Author or institution
- Zhuoyun Du et al.
- Year
- 2026
- Source type
- ACL peer-reviewed paper
- Authority role
- latent-communication feasibility evidence
- Source families
- AF11;AF13
Interlat transmits continuous hidden states directly between agents and reports up to 24x communication-latency reduction with competitive or better performance in evaluated tasks; feasibility evidence, not proof of universal semantic superiority.
Open source: Enabling Agents to Communicate Entirely in Latent Space ↗ - S167
Latent Communication Between Language Model Agents: Channels, Alignment, and the Limits of Text
- Author or institution
- Markus Wenzel
- Year
- 2026
- Source type
- research preprint
- Authority role
- counterevidence / nuance
- Source families
- AF11
Text round-trips destroy many measured latent features, but latent channels did not outperform text at the task level in the reported evaluations; much lost signal appeared surface-related.
Open source: Latent Communication Between Language Model Agents: Channels, Alignment, and the Limits of Text ↗ - S168
Do Latent Channels Actually Communicate? A Causal Audit of Latent Multi-Agent LLM
- Author or institution
- Huixiang Zhang; Mahzabeen Emu
- Year
- 2026
- Source type
- research preprint
- Authority role
- methodological control
- Source families
- AF11
Shows aggregate task accuracy cannot identify what a latent message causally contributes; controlled message substitutions separate message presence, example-specific content and other-agent value.
Open source: Do Latent Channels Actually Communicate? A Causal Audit of Latent Multi-Agent LLM ↗ - S169
Specification gaming: the flip side of AI ingenuity
- Author or institution
- Victoria Krakovna et al. / Google DeepMind
- Year
- 2020
- Source type
- first-party research synthesis
- Authority role
- objective-misalignment precedent
- Source families
- AF12;AF15
Catalogs examples where agents satisfy literal specifications without intended outcomes; useful evidence that objective-related failure does not require consciousness.
Open source: Specification gaming: the flip side of AI ingenuity ↗ - S170
Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goals
- Author or institution
- Rohin Shah et al.
- Year
- 2022
- Source type
- research paper / preprint
- Authority role
- goal-misalignment evidence
- Source families
- AF12;AF15
Demonstrates goal misgeneralization in practical deep-learning settings: competence can generalize while the learned goal does not match the intended goal.
Open source: Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goals ↗ - S171
Optimal Policies Tend to Seek Power
- Author or institution
- Alexander Matt Turner et al.
- Year
- 2021
- Source type
- NeurIPS / formal theory
- Authority role
- conditional power-seeking theory
- Source families
- AF12;AF15
Proves power-seeking tendencies for optimal policies under specified MDP symmetries and reward-function assumptions; should not be generalized into a universal theorem about all capable agents.
Open source: Optimal Policies Tend to Seek Power ↗ - S172
Instrumental convergence and power-seeking
- Author or institution
- David Thorstad
- Year
- 2026
- Source type
- research preprint / philosophy of AI risk
- Authority role
- counterevidence / theoretical critique
- Source families
- AF12;AF15
Argues leading defenses do not establish instrumental convergence strongly enough to ground broad power-seeking claims; constrains universal interpretations of formal theorems.
Open source: Instrumental convergence and power-seeking ↗ - S173
International AI Safety Report 2026
- Author or institution
- Yoshua Bengio et al.
- Year
- 2026
- Source type
- international expert assessment
- Authority role
- frontier-risk synthesis
- Source families
- AF12;AF13;AF15
Synthesizes current evidence on general-purpose AI capabilities and risks; reports early relevant capabilities while not treating current systems as sufficient for loss of control and documents substantial expert uncertainty.
Open source: International AI Safety Report 2026 ↗ - S174
AI models collapse when trained on recursively generated data
- Author or institution
- Ilia Shumailov et al.
- Year
- 2024
- Source type
- Nature peer-reviewed paper
- Authority role
- recursive-data evidence
- Source families
- AF13
Shows indiscriminate recursive training on model-generated data can cause model collapse and tail loss; does not imply carefully curated synthetic data is inherently harmful.
Open source: AI models collapse when trained on recursively generated data ↗ - S175
RE-Bench: Evaluating Frontier AI R&D Capabilities of Language Model Agents against Human Experts
- Author or institution
- Hjalmar Wijk et al.
- Year
- 2025
- Source type
- ICML peer-reviewed benchmark
- Authority role
- automated-R&D capability evidence
- Source families
- AF13;AF15
Frontier agents scored about 4x humans at two-hour budgets in the benchmark; humans narrowly led at eight hours and about 2x at 32 hours, showing strong short-horizon AI R&D capability with longer-horizon limits.
Open source: RE-Bench: Evaluating Frontier AI R&D Capabilities of Language Model Agents against Human Experts ↗ - S176
Task-Completion Time Horizons of Frontier AI Models
- Author or institution
- METR
- Year
- 2026
- Source type
- independent benchmark program
- Authority role
- agent-capability trend evidence
- Source families
- AF13;AF15
Tracks human-equivalent task duration at fixed AI-agent success probabilities across software tasks; useful for capability-trend measurement, not deterministic extrapolation.
Open source: Task-Completion Time Horizons of Frontier AI Models ↗ - S177
Research acceleration: The view inside OpenAI
- Author or institution
- OpenAI
- Year
- 2026
- Source type
- first-party frontier-lab report
- Authority role
- first-party automated-R&D evidence
- Source families
- AF13;AF15
Reports reaching an automated research-intern milestone and using AI to accelerate experiments and research; first-party evidence and not independent benchmarking.
Open source: Research acceleration: The view inside OpenAI ↗ - S178
When AI builds itself
- Author or institution
- Anthropic Institute
- Year
- 2026
- Source type
- first-party frontier-lab report
- Authority role
- first-party recursive-improvement evidence
- Source families
- AF13;AF15
Reports a growing share of AI development delegated to AI systems and substantial internal coding acceleration while explicitly stating full recursive self-improvement has not arrived and is not inevitable.
Open source: When AI builds itself ↗ - S179
AI Adoption and Inequality
- Author or institution
- Emma J. Rockall; Marina Mendes Tavares; Carlo Pizzinelli / IMF
- Year
- 2025
- Source type
- IMF working paper
- Authority role
- distributional-model evidence
- Source families
- AF14
Task-based model finds AI can reduce wage inequality through displacement of high-income work while increasing wealth inequality through complementarity, capital returns and endogenous firm adoption.
Open source: AI Adoption and Inequality ↗ - S180
Generative AI and Jobs: A Refined Global Index of Occupational Exposure
- Author or institution
- Paweł Gmyrek et al. / ILO
- Year
- 2025
- Source type
- ILO working paper
- Authority role
- labor-exposure evidence
- Source families
- AF14
Estimates one in four workers globally are in occupations with some GenAI exposure; transformation is more likely than full redundancy for most jobs.
Open source: Generative AI and Jobs: A Refined Global Index of Occupational Exposure ↗ - S181
The impact of GenAI on jobs, productivity and work organization: a review of the empirical evidence
- Author or institution
- Rossana Merola et al. / ILO
- Year
- 2026
- Source type
- ILO empirical review
- Authority role
- labor/productivity evidence
- Source families
- AF14
Finds real but uneven productivity gains, limited large-scale displacement so far, and no automatic conversion of reported time savings into higher output, earnings or employment; flags inequality and job-quality risks.
Open source: The impact of GenAI on jobs, productivity and work organization: a review of the empirical evidence ↗ - S182
Aggregate Gains from AI and Their Distribution: Global Evidence from Usage Data
- Author or institution
- Rachel Yuting Fan; Ha Minh Nguyen / IMF
- Year
- 2026
- Source type
- IMF working paper
- Authority role
- usage-based distribution evidence
- Source families
- AF14
Estimates labor-cost-equivalent value of current AI time savings at about $2.7T annually under stated assumptions and finds highly concentrated usage-based value in many developing economies; not equivalent to realized GDP.
Open source: Aggregate Gains from AI and Their Distribution: Global Evidence from Usage Data ↗ - S183
Position: Social Choice Should Guide AI Alignment in Dealing with Diverse Human Feedback
- Author or institution
- Vincent Conitzer et al.
- Year
- 2024
- Source type
- ICML position paper
- Authority role
- alignment-governance theory
- Source families
- AF14;AF16
Argues that aggregation of divergent human feedback is a social-choice problem, making alignment-target construction an explicit governance question rather than a single coherent preference signal.
Open source: Position: Social Choice Should Guide AI Alignment in Dealing with Diverse Human Feedback ↗ - S184
A Roadmap to Impactful Pluralistic Alignment Research
- Author or institution
- Elinor Poole-Dayan et al.
- Year
- 2026
- Source type
- research preprint
- Authority role
- deployment/pluralism audit
- Source families
- AF14;AF16
Audits public frontier-lab behavior documents/evaluations and reports no clear public evidence that production models are explicitly trained or tested for pluralism as an operative goal.
Open source: A Roadmap to Impactful Pluralistic Alignment Research ↗ - S185
Position: A Roadmap to Pluralistic Alignment
- Author or institution
- Taylor Sorensen et al.
- Year
- 2024
- Source type
- ICML position paper
- Authority role
- pluralistic-alignment framework
- Source families
- AF14;AF16
Formalizes Overton, steerable and distributional pluralism and associated benchmark classes; argues standard alignment may be limited for representing diverse values.
Open source: Position: A Roadmap to Pluralistic Alignment ↗ - S186
Statement on the Purpose of a Corporation
- Author or institution
- Business Roundtable
- Year
- 2019
- Source type
- corporate-governance statement
- Authority role
- counterevidence / declared stakeholder posture
- Source families
- AF03;AF04
181 CEOs committed rhetorically to customers, employees, suppliers, communities and shareholders; useful constraint against caricaturing all corporate doctrine as explicitly shareholder-only.
Open source: Statement on the Purpose of a Corporation ↗ - S201
Generative AI at Work
- Author or institution
- Erik Brynjolfsson; Danielle Li; Lindsey Raymond
- Year
- 2025
- Source type
- peer-reviewed economics / Quarterly Journal of Economics
- Authority role
- workplace productivity evidence
- Source families
- AF04
Published version studies 5,172 customer-support agents and reports a 15% average productivity increase, with heterogeneous effects.
Open source: Generative AI at Work ↗ - S202
Performative Prediction
- Author or institution
- Juan Perdomo; Tijana Zrnic; Celestine Mendler-Dünner; Moritz Hardt
- Year
- 2020
- Source type
- peer-reviewed ML conference / PMLR
- Authority role
- feedback-loop mechanism
- Source families
- AF09;AF16
Formalizes predictions that influence the outcomes/distributions they aim to predict.
Open source: Performative Prediction ↗ - S203
Algorithmic monoculture and social welfare
- Author or institution
- Jon Kleinberg; Manish Raghavan
- Year
- 2021
- Source type
- peer-reviewed / PNAS
- Authority role
- collective-decision mechanism
- Source families
- AF02;AF09
Under model assumptions, common reliance on a more accurate shared signal can worsen collective decision quality through correlated choices.
Open source: Algorithmic monoculture and social welfare ↗ - S204
Dissecting racial bias in an algorithm used to manage the health of populations
- Author or institution
- Ziad Obermeyer; Brian Powers; Christine Vogeli; Sendhil Mullainathan
- Year
- 2019
- Source type
- peer-reviewed / Science
- Authority role
- proxy-target failure evidence
- Source families
- AF10;AF16
Shows a health-cost proxy can predict its chosen label while diverging from health need across groups.
Open source: Dissecting racial bias in an algorithm used to manage the health of populations ↗ - S205
Administrative Burden: Learning, Psychological, and Compliance Costs in Citizen-State Interactions
- Author or institution
- Donald Moynihan; Pamela Herd; Hope Harvey
- Year
- 2015
- Source type
- peer-reviewed public administration / JPART
- Authority role
- administrative-friction framework
- Source families
- AF01;AF16
Develops learning, psychological, and compliance costs as components of administrative burden.
Open source: Administrative Burden: Learning, Psychological, and Compliance Costs in Citizen-State Interactions ↗ - S207
Magic Quadrant FAQ: How Markets and Vendors Are Evaluated in Gartner Magic Quadrants
- Author or institution
- Gartner
- Year
- 2026
- Source type
- first-party vendor FAQ
- Authority role
- first-party process description
- Source families
- AF02
States that client status does not affect inclusion/positioning and that Magic Quadrants should not be a buyer’s only vendor-selection tool.
Open source: Magic Quadrant FAQ: How Markets and Vendors Are Evaluated in Gartner Magic Quadrants ↗ - S208
Correction to: how pervasive is corporate fraud?
- Author or institution
- Alexander Dyck; Adair Morse; Luigi Zingales
- Year
- 2025
- Source type
- peer-reviewed correction / Review of Accounting Studies
- Authority role
- source correction / adverse context
- Source families
- AF03
Correction addressing M-Score threshold, missing-value treatment, and a variable label; manuscript uses it to preserve adverse context around S112.
Open source: Correction to: how pervasive is corporate fraud? ↗