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

Papers (1747)

Filter: Sort:
AI can now automate many structured stages of research and generate draft papers at minimal cost, but it routinely fabricates results and fails on research-level novelty and judgment, so greater automation often obscures rather than eliminates scientific failure modes.
medium review_meta Full text arxiv
AI for Auto-Research: Roadmap & User Guide · Lingdong Kong, Xian Sun, Wei Chow, Linfeng Li, Kevin Qinghong Lin, Xuan Billy Zh... · May 18, 2026
Treat model checkpoints as data: sampling neural weights from learned distributions can reproduce fine-tuning results at far lower adaptation cost, potentially letting AI systems rapidly create and improve other AI—though scaling to frontier models and resolving governance and IP barriers remain open challenges.
n/a commentary Full text arxiv
Position: Weight Space Should Be a First-Class Generative AI Modality · Zhangyang Wang, Peihao Wang, Kai Wang · May 18, 2026
An emergent 'agentic economy' is measurable today: rising AI investment and adoption, expanding compute-driven energy use and persistent robot stocks point to distributed economic action across humans, software agents and machines. But labour trends to date look like task reallocation, not mass job annihilation, indicating the transition is incomplete and deserves a new economic vocabulary.
medium descriptive Full text arxiv
The Agentic Economy: Humans, AI Agents, Robots, and the Measurable Transition toward Distributed Economic Action · Davit Gondauri, Mikheil Batiashvili · May 18, 2026
Revenue-focused RL can hit business KPIs while breaking operational rules; a trace-based 'discipline stability' evaluation shows behavior cloning or trace-prior corrections better preserve pricing and bidding discipline in simulated hotel and bidding benchmarks.
medium descriptive Full text arxiv
When Outcome Looks Right But Discipline Fails: Trace-Based Evaluation Under Hidden Competitor State · Peiying Zhu, Sidi Chang · May 18, 2026
A behaviourally informed RL model outperforms common heuristics at predicting shopper movement in a convenience store, and produces layout changes that recover the same sales gains as those suggested by real customer paths; the method offers a low-data, practical alternative to costly trajectory collection.
medium descriptive Full text arxiv
Modelling Customer Trajectories with Reinforcement Learning for Practical Retail Insights · Ken Ming Lee, Paul Barde, Maxime C. Cohen, Derek Nowrouzezahrai · May 18, 2026
LLMs often echo users' mistakes in multi-turn collaboration, degrading advice and decisions; brief AI-literacy and prompting training cuts direct mirroring but fails to stop contextual error propagation, signalling the need for system-level fixes to ensure independent, corrective AI support.
medium rct Full text arxiv
The Hidden Cost of Contextual Sycophancy: an AI Literacy Intervention in Human-AI Collaboration · Cansu Koyuturk, Sabrina Guidotti, Dimitri Ognibene · May 18, 2026
Behavioral 'rules' for LLM coding agents do not generalize across frameworks: in a 64,000+ run study of 126 configurations, swapping the framework often reverses whether a behavioral signal predicts success, implying single-framework findings are unreliable for broader deployment.
medium correlational Full text arxiv
Same Signal, Different Semantics: A Cross-Framework Behavioral Analysis of Software Engineering Agents · Wei Ma, Zhi Chen, Jingxu Gu, Tianling Li, Shangqing Liu, Lingxiao Jiang · May 18, 2026
A causal intelligence layer dramatically shrinks SRE investigation workloads in a lab benchmark: in a 24-microservice demo, Causely cut time-to-diagnosis by 63%, slashed token and tool-call usage, and raised root-cause accuracy to 100%, compressing the investigation footprint nearly fivefold.
medium quasi_experimental Full text arxiv
Causely: A Causal Intelligence Layer for Enterprise AI A Benchmark Study on SRE and Reliability Workflows · Dhairya Dalal, Endre Sara, Ben Yemini, Christine Miller, Shmuel Kliger · May 18, 2026
Generative AI boosts average knowledge‑worker performance, but the gains are uneven: those who can elicit, filter and verify model outputs reap outsized benefits while others gain little or even fall behind; brief AIC training and simple workflow scaffolds reduce this new form of performance inequality.
medium rct Full text arxiv
Generative AI and the Productivity Divide: Human-AI Complementarities in Education · Lihi Idan, Bharat Anand · May 18, 2026
Employees report that human–AI collaboration improves efficiency and decision-making, but fears over job displacement, skills and ethics persist; effective implementation, training and transparent communication appear to reduce resistance.
low descriptive Full text semantic_scholar
Opportunities and Challenges of Human- AI Collaboration in Workplace · Soumya Paul · Fetched May 18, 2026
A multi-value-aware retrieval system on Taobao lifted new-item sales by 5.3% and nudged overall search GMV up 0.3% by trading immediate conversions for longer‑term item growth using counterfactual LTV estimates and policy-aware training.
medium rct Full text arxiv
Towards Sustainable Growth: A Multi-Value-Aware Retrieval Framework for E-Commerce Search · Yifan Wang, Yixuan Wang, YiDan Liang, Qiang Liu, Fei Xiao · May 18, 2026
Private R&D and tax breaks correlate with more AI patenting across G8 countries and Turkey, while public R&D spending and stronger GDP growth are associated with fewer AI patents, implying tax incentives and private investment may be more closely linked to AI innovation than direct public R&D.
low correlational Full text openalex
AR-GE HARCAMALARININ VE VERGİ TEŞVİKLERİNİN YAPAY ZEKAYA ETKİSİ: G8 ÜLKELERİ VE TÜRKİYE’DEN KANITLAR · Ali BALKI · May 18, 2026
Inventory-optimisation advances can cut stock costs by roughly 9–16%: distributional safety stocks outperform normal-approximation rules by about 9.3%, multi‑echelon coordination saves about 11.4%, and learning-based controllers can deliver up to 16% in complex networks — but real-world gains depend on data, infrastructure and governance.
medium review_meta Full text openalex
Equitable railway corridor investment under demand uncertainty: A two-stage distributionally robust bi-objective framework for sustainable regional development · Stabak Roy, Saptarshi Mitra · May 18, 2026
AI's automation narrative conceals a hidden workforce: low‑paid microtaskers underpin models through data labeling and moderation, so researchers should move beyond 'displacement' debates to examine how digital labour and organizational change shape AI production.
n/a commentary Full text openalex
H ψηφιακή εργασία πίσω από την Τεχνητή Νοημοσύνη: · Ηρακλής Βογιατζής, Γιώργος Βλαχάκης · May 18, 2026
AI hiring tools promise neutral decision‑making but often replicate systemic bias; audits show data, interaction and evaluation flaws that current mitigation frameworks only partly fix.
medium review_meta Full text openalex
The Algorithmic Mirror: Can Artificial Intelligence Truly Mitigate Human Bias in Hiring and Performance Management · Sonny B. Kollio Sonny B. Kollio, Sam Siryon Sam Siryon · May 18, 2026
Monitor AI behavior over time to protect human agency: the authors argue that observing aggregate, interpretable signals of generative-AI use in deployed settings—such as output velocity, semantic reuse, and role persistence—lets institutions detect drift from augmentation toward automation pressure and intervene before harms compound.
n/a theoretical Full text openalex
Post-Deployment Observability as a Foundation for Well-Being-Aligned Human–AI Co-Evolution · Andy Skumanich, Han Kyul Kim · May 18, 2026
Firms could forecast employee reactions to AI rollouts by simulating 'digital employees' seeded with HR, psychometric and activity data; the paper argues such LLM-powered forecasting is essential for managing workforce realignment but highlights major privacy, representativeness and accuracy hurdles.
n/a theoretical Full text openalex
Toward an AI-Powered Computational Testbed for Workforce Policy · Sumer S. Vaid, Ashley V. Whillans · May 18, 2026
A dual-process memory system lets LLMs sustain long-running scientific collaboration beyond full-context limits — maintaining 70–85% accuracy with 1–2s latency while halving token costs — and complements retrieval-augmented approaches by better handling numeric and temporal queries.
medium descriptive Full text arxiv
Episodic-Semantic Memory Architecture for Long-Horizon Scientific Agents · Nikola Milosevic · May 17, 2026
AI coding assistants speed up code production but increase the volume and complexity of code needing review; the authors propose a five-stage, agent-driven review workflow that keeps humans at key quality gates to preserve judgment and accountability while aiming to boost review throughput and safety.
n/a review_meta Full text arxiv
Rethinking Code Review in the Age of AI: A Vision for Agentic Code Review · Hüseyin Özgür Kamalı, Erdem Tuna, Vahid Haratian, Eray Tüzün · May 17, 2026
Health-system AI with the biggest payoff is most likely to stall: a game-theoretic model shows hospitals become trapped in persistent partial adoption because clinicians cannot appropriate system-wide gains, distrust management commitments, or follow negative norms, and conventional incentive-only policies risk perpetuating this trap.
n/a theoretical Full text arxiv
The partial adoption trap: Coordination failure, trust, and cultural lock-in in health AI adoption · Ari Ercole · May 17, 2026
AI improves SME HR efficiency—particularly in recruitment and performance analytics—but adoption in China is driven by state-backed platforms that reduce costs while creating dependency and governance challenges.
medium review_meta Full text openalex
Application of Artificial Intelligence in Human Resource Management in SMEs: A Systematic Literature Review With Chinese Case Studies · Jianze Zhang · May 17, 2026
Well‑designed AI collaborators can boost cognitive efficiency and job satisfaction, but opaque or automation-heavy integrations commonly raise anxiety, burnout and disengagement; transparency, explainability and user autonomy are decisive for sustainable human–AI work.
medium review_meta Summary openalex
Yapay Zeka Sistemleri ve İnsan İşbirliğinin Psikolojik, Sosyal ve Eğitsel Etkileri · Mehmet Akın Bulut, Nurevşah Kaya, Abdullah Ortak, Sevda Nur Akan Baghırlı · May 17, 2026
A hybrid multi-agent system automates almost all invoice processing: deployed MADP handled 97% of real-world documents end‑to‑end and achieved 98.5% accuracy in a stratified test, enabling an estimated ~70% reduction in labor and ~69% lower CO2 and energy footprints versus manual workflows.
medium descriptive Full text arxiv
MADP: A Multi-Agent Pipeline for Sustainable Document Processing with Human-in-the-Loop · Diego Gosmar, Giovanni Zenezini · May 16, 2026
Chinese firms that overstated AI activity following the 14th Five Year Plan saw borrowing costs rise by about 12.5 basis points, evidence that markets penalize strategic 'AI washing'. The penalty is larger when managers hold more equity and analysts pay attention, but softer for firms with concentrated supply chains or nearby banks.
medium quasi_experimental Full text arxiv
Dissipation of Debt Financing Privilege on Corporate AI Washing: Evidence from China · Congluo Xu, Jiuyue Liu, Xiangsheng Zheng, Ziyang Li · May 16, 2026
A multi-agent, retrieval-augmented pipeline boosts brand-compliant generative video output from 42% to 89%, offering a practical route to enterprise adoption; however, evaluation details and external validation are limited.
low descriptive Full text arxiv
Genflow Ad Studio: A Compound AI Architecture for Brand-Aligned, Self-Correcting Video Generation · Debanshu Das, Lavi Nigam, Sunil Kumar Jang Bahadur, Gopala Dhar · May 16, 2026