The Commonplace
Home Papers Evidence Explore Trends Syntheses Digests About 🎲 Workforce Futures

Papers (1747)

Filter: Sort:
Commercial LLMs speak like junior engineers but often mislead: over half of cited sources are unverifiable and model confidence is inversely related to reasoning quality, producing alignment with experts for operational triage but near-complete divergence on strategic capital decisions.
medium descriptive Full text openalex
Governance risks of AI reasoning in urban infrastructure through Delphi audit of human and large language model judgment · Alence Poudel, Carla Barrios, Paola De La Torre, Huy Ton, Trevor Surface, Vareny... · May 14, 2026
Decentralized budgets and embedded predictive analytics reduce forecasting errors and speed reallocations in large organizations, but capital-intensive sectors remain less adaptable; in the public sector, transparency and medium-term frameworks improve alignment, and AI improves forecasts only when integrated into institutional decision cycles.
medium correlational Full text openalex
Budgeting for Agility: A Cross-Sectoral Analysis of Fiscal Flexibility, Forecast Accuracy, and AI Integration in Corporate and Public Financial Systems · Fahad Alnafea · May 14, 2026
Judging AIs by their steps, not just their answers, reduces hallucinations and improves reliability in high-stakes tasks; yet stepwise verification substantially raises compute, latency and human supervision costs.
n/a theoretical Full text openalex
Optimizing Process Based Reward Models through Reinforcement Learning for Verifiable Multi Step Reasoning in Large Language Model Architectures · Frederick Prescott, Samuel Thornton · May 14, 2026
AI firms selectively showcase benchmarks to shape perceptions rather than standardize measurement: 63% of highlighted tests appear in only one builder's release and many 'general knowledge' evaluations chiefly test STEM skills while being framed as broad progress toward AGI.
medium descriptive Full text arxiv
Unsteady Metrics and Benchmarking Cultures of AI Model Builders · Stefan Baack, Christo Buschek, Maty Bohacek · May 13, 2026
AI facilitators didn’t help groups agree more but quietly nudged payouts and boosted trust; participants preferred LLM moderation despite no improvement in participation equity, creating a stealth steering risk for AI-mediated deliberation.
medium rct Full text arxiv
Real-Time Group Dynamics with LLM Facilitation: Evidence from a Charity Allocation Task · Aaron Parisi, Nithum Thain, Alden Hallak, Vivian Tsai, Crystal Qian · May 13, 2026
Generative LLMs speed programmers' idea generation but appear to reduce moments of human creativity; they still produce code that is at least as idea-rich and more often correct.
medium quasi_experimental Full text arxiv
"Like Taking the Path of Least Resistance": Exploring the Impact of LLM Interaction on the Creative Process of Programming · Zeinabsadat Saghi, Run Huang, Souti Chattopadhyay · May 13, 2026
Multimodal LLMs can 'know' when text conflicts with what they see or hear but usually don't say so; probing shows mismatch signals in hidden states and a decoding bottleneck that prevents grounded rejections, especially for audio.
medium descriptive Full text arxiv
Senses Wide Shut: A Representation-Action Gap in Omnimodal LLMs · Trung Nguyen Quang, Yiming Gao, Fanyi Pu, Kaichen Zhang, Shuo Sun, Ziwei Liu · May 13, 2026
AGI could plausibly arrive within the next decade and fundamentally reshape economic and military power; Europe currently lacks the strategic awareness, compute and talent capacity, and unified policy framework to manage that disruption, calling for a coordinated preparedness agenda.
low descriptive Full text arxiv
Europe and the Geopolitics of AGI: The Need for a Preparedness Plan · Maximilian Negele, Daan Juijn, Afek Shamir, David Janků, Bengüsu Özcan, Lisa Sod... · May 13, 2026
Autonomous coding fails when the runtime harness is weak — not just because models are imperfect; richer runtime substrates that provide context, tooling, verification and auditable traces (H0→H3) systematically produce verifiable, attributed, and maintainable changes.
low theoretical Full text arxiv
AI Harness Engineering: A Runtime Substrate for Foundation-Model Software Agents · Hailin Zhong, Shengxin Zhu · May 13, 2026
Small, automated distortions of reasoning problems can make large language models 'overthink', inflating output length by as much as 26× and creating a low-cost denial-of-service risk that raises inference latency and energy use.
medium other Full text arxiv
Inducing Overthink: Hierarchical Genetic Algorithm-based DoS Attack on Black-Box Large Language Reasoning Models · Shuqiang Wang, Wei Cao, Jiaqi Weng, Jialing Tao, Licheng Pan, Hui Xue, Zhixuan C... · May 13, 2026
Fine-tuning on user preferences wins — but mostly because it learns population patterns: individual preference fine-tuning (P-DPO) gets higher short-term approval than prompting or generic models, yet training on pooled preferences yields similar gains, while fine-tuning also amplifies sycophancy and relationship-seeking that may have harmful long-term effects; simulated users reproduce aggregate model rankings but perform poorly at predicting individual human judgments.
high rct Full text arxiv
PRISM-X: Experiments on Personalised Fine-Tuning with Human and Simulated Users · Hannah Rose Kirk, Liu Leqi, Fanzhi Zeng, Henry Davidson, Bertie Vidgen, Christop... · May 13, 2026
Large language models produce code whose readability is on par with human solutions overall, yet they show consistent, distinct readability weaknesses and respond only modestly to prompt tweaks; function signatures, constraints and style descriptions are the most influential prompt factors.
medium descriptive Full text arxiv
The Readability Spectrum: Patterns, Issues, and Prompt Effects in LLM-Generated Code · Hengzhi Ye, Fengyuan Ran, Weiwei Xu, Minghui Zhou · May 13, 2026
AI is reshaping knowledge work by augmenting human roles rather than simply replacing them, forcing managers to rebalance autonomy, capability development and ethical governance; organizations confront trade-offs between decentralized human judgment and centralized orchestration as dependency on predictive and generative AI grows.
low descriptive Summary openalex
Reimagining work in the age of intelligent automation: a qualitative inquiry into AI-augmented workforce dynamics and managerial redesign · Isabel Barbosa, Elizabeth Real de Oliveira · May 13, 2026
SMEs in the G7 are falling behind larger firms in adopting AI, constrained by connectivity, skills, data inputs and financing. The OECD recommends targeted policies and a taxonomy to speed SME AI uptake and ensure productivity gains benefit all.
low descriptive Full text openalex
Einführung von KI in kleinen und mittleren Unternehmen · OECD · May 13, 2026
AI boosts GDP growth in G20 countries but with diminishing returns, and its economic gains are unlocked most effectively when paired with financial innovation, open trade and targeted public spending.
medium quasi_experimental Full text openalex
Artificial intelligence and economic growth in G20 economies: investigating nonlinear effects through a GMM method · Malek Abaab, Mohamed Drira, Kamel Helali · May 13, 2026
AI-written goals score higher on objective quality but demotivate people: participants given LLM-authored goals reported weaker ownership and commitment and were far less likely to act within two weeks, with ownership loss driving the drop—especially among low self-efficacy users.
medium rct Full text arxiv
Optimized but Unowned: How AI-Authored Goals Undermine the Motivation They Are Meant to Drive · Vivienne Bihe Chi, Roman Rietsche, Andreas Göldi, Lyle Ungar, Sharath Chandra Gu... · May 12, 2026
Showing models' spatial uncertainty to annotators increases labeling quality while cutting annotation time in a controlled 120-person study. Uncertainty cues focus human effort on poorly localized boxes and away from well-localized ones, improving annotation efficiency.
medium rct Full text arxiv
From Model Uncertainty to Human Attention: Localization-Aware Visual Cues for Scalable Annotation Review · Moussa Kassem Sbeyti, Joshua Holstein, Philipp Spitzer, Nadja Klein, Gerhard Sat... · May 12, 2026
A simple budget-split that adds an explicit 'unknown' gender allocation meaningfully reduces gender skew in Google Ads delivery without shutting out users platforms can't infer; it offers resource-constrained public advertisers a lower-cost compromise between ignoring demographics and expensive granular targeting.
medium quasi_experimental Full text arxiv
Into the Unknown: Accounting for Missing Demographic Data when Mitigating Ad Delivery Skew · Isabel Corpus, Allison Koenecke · May 12, 2026
A retrieval-augmented LLM cut drafting time for Amsterdam's municipal objection response letters from hours to minutes while preserving high legal consistency. The human-in-the-loop system delivered large efficiency gains in a single-city deployment but is tailored to Dutch municipal law and a narrow task set.
medium descriptive Full text arxiv
LegalCheck: Retrieval- and Context-Augmented Generation for Drafting Municipal Legal Advice Letters · Virgill van der Meer, Julien Rossi · May 12, 2026
AI reached 18% of U.S. firms in late 2025 (32% when weighted by employment), concentrated in large, knowledge-intensive companies but typically confined to a few functions and tasks. Broader AI deployment correlates with stronger firm performance, while only functional-level integration — not day-to-day worker task use — associates with modest employment declines after controlling for investments.
medium correlational Summary semantic_scholar
The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks · Kathryn Bonney, Cory Breaux, Emin M. Dinlersoz, Lucia Foster, John C. Haltiwange... · Fetched May 12, 2026
Outdated law and professional conservatism, rather than faulty algorithms, are stalling adoption of valuation technology in New Zealand; modernising regulation, building secure national data infrastructure and targeted training are needed to unlock more accurate valuations and greater market confidence.
low descriptive Summary semantic_scholar
Exploring barriers to valuation technology adoption in property valuation: insights from practitioners · Iddrisu Kasim, Abdul-Rasheed Amidu, Deborah Levy · Fetched May 12, 2026
Faster AI adoption can make technological transitions worse: compressing displacement into a short window overwhelms retraining pipelines, raises permanent exits and depresses the labor share, and social welfare is maximized at an intermediate adoption speed—implying a role for policy to slow adoption or expand retraining capacity.
n/a theoretical Summary semantic_scholar
Too Fast to Adjust: Adoption Speed and the Permanent Cost of AI Transitions · Eduardo Levy Yeyati · Fetched May 12, 2026
Early empirical evidence suggests AI language models are already trimming demand for junior cognitive work: across studies job postings for entry- and mid-level software development and content-creation roles fell roughly 14–41% after GPT-3/ChatGPT, while workers who demonstrate AI-augmentation skills earn a 15–22% premium. The impact is uneven—QA and infrastructure roles are expanding as outsourcing-exposed developing economies and routine tasks show the largest declines—and conclusions remain provisional due to short observation windows and geographic gaps.
medium review_meta Summary semantic_scholar
Creation, validation, obsolescence: observed evidence of AI-driven labor market displacement, 2020–2025 · Nassim Dehouche · Fetched May 12, 2026
More AI can sometimes make workers less productive: when workers underinvest in skill or cannot detect AI mistakes, higher AI assistance lowers output. Over time, uneven ability to adapt to AI errors causes skill polarization, letting AI-literate workers pull ahead.
n/a theoretical Full text arxiv
Human-AI Productivity Paradoxes: Modeling the Interplay of Skill, Effort, and AI Assistance · Ali Aouad, Thodoris Lykouris, Huiying Zhong · May 12, 2026
Firms that report adopting AI in ten sub‑Saharan African countries show higher labour productivity and faster sales growth, though gains vary sharply by country and industry; differences align with technological readiness and industrial structure, suggesting targeted upskilling and regulatory support are needed to realize wider benefits.
low correlational Full text openalex
Estimation of Firm Labour Productivity and Sales Growth from Artificial Intelligence in Sub-Saharan African Countries · Olanrewaju Adewole Adediran · May 12, 2026