The Commonplace
Home Dashboard Papers Evidence Syntheses Digests 🎲

Papers (1134)

Filter: Sort:
A five‑node value chain and 'Innovation Frontier' map how AI is reshaping legal work, arguing that targeted oversight can contain professional and systemic risks while enabling broader access and quality gains in legal services.
n/a commentary openalex
Rewired: Reconceptualizing Legal Services for the AI Age · Lev Breydo · Apr 01, 2026
AI adoption initially boosts firms' electricity demand—suggesting energy needs outpace early efficiency gains—but the spike fades and is gone after about three years, with larger short-run increases in advanced regions, competitive industries, manufacturing firms, small and non-state firms, and low-tech or low-pollution firms.
medium quasi_experimental openalex
The Impact of AI Adoption on Electricity Output Growth Gap: Evidence from Listed Chinese Firms · Guoyao Wu, Zhiqiang Lan, Yang Xu, Ye Guo · Apr 01, 2026
China's AI pilot zones measurably boost manufacturers' operational resilience: firms in designated AIIAPZ areas show significant gains in resilience driven by lower internal agency frictions and better supply‑chain allocation, with largest effects for coastal, growth-stage, tech- and capital‑intensive firms.
medium quasi_experimental openalex
Does Artificial Intelligence Improve the Operational Resilience of Enterprises? Evidence from the AI Innovative Application Pioneer Zone Policy in China · Yiting Hu, Xu Yan, Chaofan Duan, Xiaodong Yang, Jiaoping Yang · Apr 01, 2026
Robotics manufacturing first raises then lowers urban carbon emissions in China: expansion initially boosts emissions, but once the industry attains moderate scale—especially via system integration—wider robot deployment and efficiency gains cut emissions, with stronger mitigation in central regions than in the east.
medium correlational openalex
Exploring the nonlinear relationship between robotics manufacturing and urban carbon emissions · Jie Lin, Yizhi Xie, Jianfu Shen · Apr 01, 2026
Generative AI can quietly hollow out public knowledge archives by resolving problems privately and discouraging contributors, creating self-reinforcing low-archive traps; nudging users to share AI-assisted solutions can offset private diversion but cannot restore participation-driven declines without policies to sustain contributors.
n/a theoretical openalex
When AI Improves Answers but Slows Knowledge Creation: Matching and Dynamic Knowledge Creation in Digital Public Goods · Keh-Kuan Sun · Apr 01, 2026
Agentic AI could put nearly all information-intensive occupations in major U.S. tech hubs at moderate displacement risk by 2030: a new ATE metric predicts 93% of 236 occupations in five regions cross the moderate-risk threshold, though the result rests on calibrated capability and adoption assumptions; the model also forecasts seventeen emerging human-AI and governance roles.
low theoretical arxiv
Agentic AI and Occupational Displacement: A Multi-Regional Task Exposure Analysis of Emerging Labor Market Disruption · Ravish Gupta, Saket Kumar · Mar 31, 2026
Top-tier commercial LLMs can mimic official legislative reasoning for routine bills, while open-weight models lag by a full tier; however, all models occasionally fabricate plausible but unsupported rationales on novel or politically unusual proposals, creating a risk of contextual ignorance rather than stable bias.
low descriptive arxiv
Can Commercial LLMs Be Parliamentary Political Companions? Comparing LLM Reasoning Against Romanian Legislative Expuneri de Motive · Iulian Lucău, Adelin-George Voicu · Mar 31, 2026
A governance gap — not just automation — threatens to lock in unequal, opaque AI decision‑making as labor is displaced; without redesigning oversight to ensure real cognitive, technical and institutional access, societies risk a 10–15 year window of path‑dependent harm.
n/a commentary arxiv
Beyond Symbolic Control: Societal Consequences of AI-Driven Workforce Displacement and the Imperative for Genuine Human Oversight Architectures · Richard J. Mitchell · Mar 31, 2026
Structured intent prompts (5W3H/CO-STAR/RISEN) make AI understanding far more consistent across languages and models and speed up human–AI interactions, with the biggest improvements on weaker models; users needed 60% fewer rounds and reported much higher satisfaction when AI-expanded 5W3H prompts were used.
medium quasi_experimental arxiv
Structured Intent as a Protocol-Like Communication Layer: Cross-Model Robustness, Framework Comparison, and the Weak-Model Compensation Effect · Peng Gang · Mar 31, 2026
Keeping human and AI judgments independent and resolving disagreements with a second human consistently beats the common AI-advisor model across ten tasks, from medical diagnosis to misinformation detection. The gain stems from people’s poor ability to distinguish correct from incorrect AI advice, so a simple tie-breaking hybrid design improves accuracy and transparency.
medium rct arxiv
Beyond AI advice -- independent aggregation boosts human-AI accuracy · Julian Berger, Pantelis P. Analytis, Ville Satopää, Ralf H. J. M. Kurvers · Mar 31, 2026
A wave of rapid technology creation from the 1970s through the 1990s lifted the college wage premium by roughly 30% by privileging workers who learn new technologies faster; as technologies aged and standardized the effect faded, explaining the premium's flattening after 2010.
medium theoretical manual
THE SKILL PREMIUM IN TIMES OF RAPID TECHNOLOGICAL CHANGE · · Fetched Mar 31, 2026
Policymakers can steer AI development to protect jobs and wages: a planner who cares about workers' welfare will favor labor-complementary and capital-augmenting innovations and may tax automation. Yet if AI sufficiently devalues labor, steering becomes less effective and policy should instead prioritize redistribution and improving non-monetary aspects of well-being.
n/a theoretical manual
NBER WORKING PAPER SERIES · · Fetched Mar 31, 2026
Large language models can boost short-term output while eroding users’ ability to judge their own knowledge; a new 'AI-mediated metacognitive decoupling' model explains why people become overconfident, over- or under-rely on tools, and transfer skills weakly.
medium review_meta arxiv
Beyond the Steeper Curve: AI-Mediated Metacognitive Decoupling and the Limits of the Dunning-Kruger Metaphor · Christopher Koch · Mar 31, 2026
Households that adopt ChatGPT reallocate saved time into online leisure rather than increasing productive browsing, implying generative AI sharply raises the efficiency of household digital tasks. Adoption has been fastest among younger and higher-income households, producing a widening 'generative AI divide' in who benefits.
medium quasi_experimental manual
https://arxiv.org/pdf/2603.03144 · · Fetched Mar 31, 2026
An autonomous evolutionary system, ASI-Evolve, discovered hundreds of improved models and pipelines—finding 105 state-of-the-art linear-attention architectures and yielding large benchmark gains (e.g., +18 MMLU, up to +12.5 RL points)—offering early evidence that AI can speed parts of the AI research loop.
medium other arxiv
ASI-Evolve: AI Accelerates AI · Weixian Xu, Tiantian Mi, Yixiu Liu, Yang Nan, Zhimeng Zhou, Lyumanshan Ye, Lin Z... · Mar 31, 2026
Parallel subagents act like resilient search engines, excelling at fast, broad automated optimization under tight time limits; specialist agent teams are better suited to deep architectural refactorings but are costlier and more fragile, needing larger compute budgets to pay off.
medium descriptive arxiv
An Empirical Study of Multi-Agent Collaboration for Automated Research · Yang Shen, Zhenyi Yi, Ziyi Zhao, Lijun Sun, Dongyang Li, Chin-Teng Lin, Yuhui Sh... · Mar 31, 2026
Subsidies can be a race to the bottom when network effects are weak or products are close substitutes; but with strong network externalities and differentiated products, competing industrial policy can create mutual gains—especially when it spurs product innovation rather than process improvements.
n/a theoretical arxiv
Industrial Policy with Network Externalities: Race to the Bottom vs. Win-Win Outcome · Nigar Hashimzade, Haoran Sun · Mar 31, 2026
Most active social-media users report feeds that echo their own views, and greater selective exposure correlates with stronger political polarization; users' perceptions of algorithmic curation amplify these associations, suggesting platform design and user behaviour jointly reinforce ideological divides.
low correlational semantic_scholar
Echo Chambers, Filter Bubbles, and Selective Exposure: Media Use and Opinion Formation in Polarized Digital Spaces · M. Hameed, Muhammad Faheem, Durdana, Aqib Hussain · Fetched Mar 31, 2026
Researchers use LLM research assistants more like collaborators than search engines, submitting longer, more complex queries and delegating tasks such as drafting and gap identification; experienced users become more targeted and engage citations more deeply, though simple keyword queries persist.
n/a descriptive semantic_scholar
Understanding Usage and Engagement in AI-Powered Scientific Research Tools: The Asta Interaction Dataset · Dany Haddad, Daniel Bareket, Joseph Chee Chang, Jay DeYoung, Jena D. Hwang, Uri ... · Fetched Mar 31, 2026
AI's effect on equity markets is ambiguous: productivity gains lift corporate cash flows, but worker displacement and systemic alignment risk can push required returns higher; in deep markets alignment-related tail risk dominates, while in shallow markets withdrawal of retail investors produces persistent elevation of the equity risk premium.
low theoretical semantic_scholar
When Does AI Raise the Equity Risk Premium? Displacement, Participation, and Structural Regimes · Rajan Raju · Fetched Mar 31, 2026
Improved LLMs and a systematic audit reveal ELT automation is far further along than ELT-Bench originally suggested: extraction and loading are largely solved and many supposed transformation failures reflected benchmark mistakes rather than model limitations.
medium descriptive arxiv
ELT-Bench-Verified: Benchmark Quality Issues Underestimate AI Agent Capabilities · Christopher Zanoli, Andrea Giovannini, Tengjun Jin, Ana Klimovic, Yotam Perlitz · Mar 31, 2026
AI that learns from people’s choices predicts risk preferences better than AI fed written instructions, but humans frequently choose or defer to prompts nonetheless — and when prompts and choices disagree, agents often follow the less accurate prompt.
medium quasi_experimental arxiv
Should I State or Should I Show? Aligning AI with Human Preferences · Keaton Ellis, Wanying Huang · Mar 31, 2026
Because AI performance improves with diminishing returns to data, compute, and model size, near-perfect accuracy is disproportionately costly, so firms often choose partial human-AI collaboration rather than full automation; calibrated to computer vision, cost-effective automation covers roughly 11% of exposed wages at the firm level but can scale far higher when AI services spread fixed costs across users.
medium theoretical arxiv
Economics of Human and AI Collaboration: When is Partial Automation More Attractive than Full Automation? · Wensu Li, Atin Aboutorabi, Harry Lyu, Kaizhi Qian, Martin Fleming, Brian C. Goeh... · Mar 31, 2026
AI and robotics are renewing growth and productivity across sectors, but gains are uneven and hinge on policy: without strategic regulation, upskilling and data infrastructure, economic benefits risk being concentrated and fragile.
low review_meta openalex
AI and Robotics Redefine Output and Growth: The New Productivity Equation · Ms. Aaradhya Nagyan · Mar 31, 2026
AI is boosting accounting efficiency and accuracy by automating repetitive tasks and improving fraud detection, while largely complementing rather than replacing accountants; adoption and impact are constrained by data, interpretability, skills and regulatory gaps.
low review_meta openalex
Role of Artificial Intelligence in the Accounting Sector · Kishor Rajendra Adhav · Mar 31, 2026