The Commonplace
Home Dashboard Papers Evidence Syntheses Digests 🎲

Papers (1134)

Filter: Sort:
A Sino‑Russian 'semi‑core' offers an epistemological as well as geopolitical challenge to Western order, elevating technical AI workers from technicians to strategic actors; their position between competing infrastructures gives them unique leverage to influence the direction of the current global interregnum.
low theoretical openalex
Theorising the Interregnum: · Stephen Wilkinson · Apr 04, 2026
AI can raise short-term output while quietly eroding human skills, and even farsighted managers may rationally accept a lower long-run productivity once AI is adopted; short-termist incentives or external value of skills can trap workers in harmful deskilling paths.
n/a theoretical arxiv
The Augmentation Trap: AI Productivity and the Cost of Cognitive Offloading · Michael Caosun, Sinan Aral · Apr 03, 2026
AI coding models generated thousands of reliable unit tests in hours and substantially reduced regression risk during major refactors; supervised, iterative test generation constrained model-driven code changes and cut manual labor for a prototype codebase.
low descriptive arxiv
AI-Assisted Unit Test Writing and Test-Driven Code Refactoring: A Case Study · Ema Smolic, Mario Brcic, Luka Hobor, Mihael Kovac · Apr 03, 2026
A survey of 627 studies identifies four distinct AI–human decision-making paradigms driven by AI–human dynamics and decision typologies; the framework helps organizations choose between intuitive, algorithmic, analytical and hybrid decision modes as they integrate AI into operations.
n/a review_meta openalex
Advancing Decision-Making through AI-Human Collaboration: A Systematic Review and Conceptual Framework · Han Li, Feng Tian · Apr 03, 2026
A hybrid local–global credit multi-agent RL cuts simulated supply-chain costs by about 26% and boosts service levels by roughly 43%, while retaining robustness under multiple simultaneous disruptions and scaling to 120 nodes.
low descriptive openalex
Research on Multi-Agent Collaborative Decision-Making Algorithm for Supply Chain Management · Changgeng Li, Zixi Liu · Apr 03, 2026
Training models to solve multiple problems in one context slashes per-problem token consumption by up to 62.6% while preserving accuracy, promising substantial inference-cost savings; the simple BCR structural change also avoids instability associated with explicit length penalties.
medium descriptive arxiv
Batched Contextual Reinforcement: A Task-Scaling Law for Efficient Reasoning · Bangji Yang, Hongbo Ma, Jiajun Fan, Ge Liu · Apr 02, 2026
AI widens firms' reach into distant knowledge but can hollow out radical discovery once automation is pervasive; beyond a threshold of AI automation firms shift from riskier, radical recombinations to safer incremental tweaks, and full automation risks eliminating originality altogether.
n/a theoretical arxiv
Bridging Distant Ideas: the Impact of AI on R&D and Recombinant Innovation · Emanuele Bazzichi, Massimo Riccaboni, Fulvio Castellacci · Apr 02, 2026
Adaptive, GPT-powered life-insurance questionnaires cut the number of questions and win user preference, but conventional forms still slightly outperform them on risk-assessment accuracy in two field experiments; further development could close the accuracy gap and streamline underwriting.
medium quasi_experimental arxiv
AI in Insurance: Adaptive Questionnaires for Improved Risk Profiling · Diogo Silva, João Teixeira, Bruno Lima · Apr 02, 2026
A prototype payment gateway lets autonomous agents spend fiat while enforcing programmatic spend policies: lab tests show policy rules cut total spending by 27.3% and security checks blocked all replay and token misuse with ~20ms latency overhead.
medium descriptive arxiv
APEX: Agent Payment Execution with Policy for Autonomous Agent API Access · Mohd Safwan Uddin, Mohammed Mouzam, Mohammed Imran, Syed Badar Uddin Faizan · Apr 02, 2026
Large language models give fairly accurate point estimates but are wildly overconfident — their nominal 95% credible intervals contain the truth only 9–44% of the time; a simple statistical recalibration (conformal prediction) expands intervals to recover the intended coverage.
medium descriptive arxiv
Bayesian Elicitation with LLMs: Model Size Helps, Extra "Reasoning" Doesn't Always · Luka Hobor, Mario Brcic, Mihael Kovac, Kristijan Poje · Apr 02, 2026
AI-made videos match humans in total attention by flooding the feed: AIGC creators win aggregate engagement through sheer volume even though viewers prefer human-made clips, and the platform’s recommendation algorithm largely determines how much AI content viewers actually see — suggesting platforms should adopt AIGC-aware distribution and governance rules.
medium correlational arxiv
Scale over Preference: The Impact of AI-Generated Content on Online Content Ecology · Tianhao Shi, Yang Zhang, Xiaoyan Zhao, Fengbin Zhu, Chenyi Lei, Han Li, Wenwu Ou... · Apr 02, 2026
A role-separated multi-agent LLM system cut curation effort to a single operator completing campus-scale datasets in two days and stopped a coordinate-transformation error affecting 2,452 stations before publication; audited handoffs detected the issue within 10 minutes and resolved it in 80 minutes.
medium descriptive arxiv
Exploring Robust Multi-Agent Workflows for Environmental Data Management · Boyuan Guan, Jason Liu, Yanzhao Wu, Kiavash Bahreini · Apr 02, 2026
A new industrial PHM benchmark finds leading LLM agents complete just 68% of realistic diagnostics and prognostics tasks, plagued by tool-orchestration errors and poor generalization; top configurations show 23% incorrect sequencing, a 14.9 percentage-point drop on multi-asset reasoning, and only 42.7% success on held-out datasets.
medium descriptive arxiv
PHMForge: A Scenario-Driven Agentic Benchmark for Industrial Asset Lifecycle Maintenance · Ayan Das, Dhaval Patel · Apr 02, 2026
A benchmark built from real production coding sessions finds current foundation models solve roughly 53–72% of fail-to-pass tasks, and those that execute tests and static analysis perform substantially better, suggesting that exposing codebase-specific verification tools materially improves externally trained agents' effectiveness.
medium descriptive arxiv
ProdCodeBench: A Production-Derived Benchmark for Evaluating AI Coding Agents · Smriti Jha, Matteo Paltenghi, Chandra Maddila, Vijayaraghavan Murali, Shubham Ug... · Apr 02, 2026
Chinese manufacturers that invest in AI show materially lower carbon intensity, largely by reshaping governance, decarbonizing production and redirecting investment to green assets; the emission gains are concentrated in firms with dispersed supply chains, in environmentally sensitive industries and in regions with underdeveloped factor markets.
medium correlational openalex
Artificial Intelligence Innovation, Internal Structure Optimization and Corporate Carbon Emission Reduction: Experience from China · Xingxing Lu, Lianying Liao, Xiaojuan Luo, Bing Zhao · Apr 02, 2026
Democratic decline and concentrated AI capability risk producing a new neo‑feudal order: a tiny class of infrastructure owners could capture economic value and coercive power, leaving most people politically and economically disenfranchised; standard policy fixes such as UBI may serve as pacification unless confronted by redistributive pressure.
low theoretical openalex
The Great Compression: Geopolitical Fragmentation, AI, and the Coming Neo-Feudal Order · Priyanshu Jain · Apr 02, 2026
Multimodal GeoAI can diagnose transport inequalities across cities but current studies are thin, uneven and poorly validated; the field needs equity-aware models, participatory workflows and stronger data governance such as urban data trusts to translate technical advances into inclusive mobility outcomes.
n/a review_meta openalex
GeoAI and Multimodal Geospatial Data Fusion for Inclusive Urban Mobility: Methods, Applications, and Future Directions · Atakilti Brhanu Kiros, Y. Ribakov, Israel Klein, Achituv Cohen · Apr 02, 2026
Organizing work around people — not just buying technology — unlocks value from automation: the authors show theoretically and document in Colombian firms that better workplace design amplifies returns to technology (management quality raises the marginal return by about 30%), and they introduce a 36‑item index to diagnose firms' human-centric readiness.
medium theoretical arxiv
From Automation to Augmentation: A Framework for Designing Human-Centric Work Environments in Society 5.0 · Cristian Espinal Maya · Apr 01, 2026
Large language models are broadly getting better across text tasks rather than suddenly mastering narrow task clusters: measured success rose from roughly 50% in mid‑2024 to 65% by mid‑2025 and is projected to reach 80–95% for most text tasks by 2029. That suggests gradual, economy‑wide automation (‘rising tide’) rather than imminent, concentrated disruptions, although actual labor impacts will depend on organizational adoption and quality requirements.
medium descriptive arxiv
Crashing Waves vs. Rising Tides: Preliminary Findings on AI Automation from Thousands of Worker Evaluations of Labor Market Tasks · Matthias Mertens, Adam Kuzee, Brittany S. Harris, Harry Lyu, Wensu Li, Jonathan ... · Apr 01, 2026
AI raises wages for workers whose tasks are augmentable — but only in the formal sector; informal workers do not capture augmentation rents, implying labor market institutions, not technology access, are the binding constraint in the Global South.
medium correlational arxiv
Augmented Human Capital: A Unified Theory and LLM-Based Measurement Framework for Cognitive Factor Decomposition in AI-Augmented Economies · Cristian Espinal Maya · Apr 01, 2026
Autonomous coding agents are increasingly active in open-source projects, but their patches are more fragile — agent-authored pull requests are rising in frequency yet are revised or removed at higher rates than human contributions, raising maintainability concerns.
medium correlational arxiv
Investigating Autonomous Agent Contributions in the Wild: Activity Patterns and Code Change over Time · Razvan Mihai Popescu, David Gros, Andrei Botocan, Rahul Pandita, Prem Devanbu, M... · Apr 01, 2026
Generative AI can sharpen managerial performance — but only under emotionally intelligent leadership; without EI-driven trust calibration, AI-enabled delegation and communications risk amplifying errors and eroding team dynamics.
n/a theoretical semantic_scholar
LEADER EMOTIONAL INTELLIGENCE IN THE GENERATIVE AI ERA: “HUMAN–AI LEADERSHIP” AS A NEW MANAGERIAL COMPETENCE · L. Golovkova, K. Hannouf · Fetched Apr 01, 2026
Adding a three-layer enterprise ontology to LLM agents cuts hallucinations and raises compliance and role fidelity: in a 600-run controlled test across five industries, ontology-grounded agents substantially outperformed ungrounded ones on accuracy, regulatory compliance and role consistency, with the biggest benefits in domains poorly covered by the LLM's training data.
medium quasi_experimental arxiv
Ontology-Constrained Neural Reasoning in Enterprise Agentic Systems: A Neurosymbolic Architecture for Domain-Grounded AI Agents · Thanh Luong Tuan · Apr 01, 2026
Personality determines who benefits from AI coaching: a theory-driven conversational coach helps overcontrolled negotiators, resilient users improve more from a traditional handbook, and undercontrolled users gain little, suggesting AI coaching should adapt support intensity to individual readiness.
medium rct arxiv
Not My Truce: Personality Differences in AI-Mediated Workplace Negotiation · Veda Duddu, Jash Rajesh Parekh, Andy Mao, Hanyi Min, Ziang Xiao, Vedant Das Swai... · Apr 01, 2026
Code-aware conversational assistants are reshaping developer workflows: rather than giving complete instructions up front, programmers iteratively refine AI outputs, offload diagnosis and validation tasks to models, and steer AI autonomy by embedding context and constraints into ongoing chats.
medium descriptive arxiv
Programming by Chat: A Large-Scale Behavioral Analysis of 11,579 Real-World AI-Assisted IDE Sessions · Ningzhi Tang, Chaoran Chen, Zihan Fang, Gelei Xu, Maria Dhakal, Yiyu Shi, Collin... · Apr 01, 2026