How AI impacts skill formation
Judy Hanwen Shen, Alex Tamkin
This study examines the impact of AI coding assistants on the skill development of novice programmers learning a new asynchronous Python library. The research identified six distinct patterns of interaction with AI. It found that only three of these involved sufficient cognitive engagement to preserve learning outcomes, highlighting that the mode of AI assistance, rather than its mere presence, determines whether workers develop the supervisory skills necessary for effective human-AI collaboration.
Inference-time reasoning selectively reduces implicit social bias in Large Language Models
Molly Apsel, Michael N. Jones
This paper examines the impact of inference-time reasoning on implicit social bias in LLMs, drawing on the psychological distinction between explicit and implicit bias. It finds that enabling reasoning during inference significantly reduces measured implicit bias in IAT-style evaluations across fifteen stereotype topics for certain model classes. This phenomenon appears to be specific to social bias domains, as no corresponding decrease is observed for non-social implicit associations.
Using AI and Big Data to predict inflation trends based on global economic sentiment
Fariba Chowdhury, Ahmed Selim Anwar, Md Abu Sayem
This study demonstrates that combining traditional macroeconomic indicators with real-time sentiment analysis from news and social media using machine learning techniques can significantly improve the accuracy of inflation forecasting. The findings indicate that AI models enhanced with sentiment analysis can offer more accurate and precise inflation predictions than conventional methods.
Agnieszka Bieńkowska, Jacek Małecki, Alexander Mathiesen-Ohman, Katarzyna Tworek
This research presents the Human-AI Symbiosis Dynamic Capabilities (HASDC) framework, which explains how collaborative human-AI systems, structured to maintain cognitive uncertainty and allow for ambiguity, generate value and evolve in unpredictable contexts. Rather than merely recognising AI as an instrument, the framework conceptualises it as an intellectual collaborator that develops alongside human interpretation processes.
AI skills improve job prospects: causal evidence from a hiring experiment
Fabian Stephany, Ole Teutloff, Angelo Leone
This study explores how AI-related skills can enhance employability. Based on a paired conjoint experiment involving 1,700 recruiters in the UK and US, the research shows that possessing AI competencies can offset disadvantages such as older age or a lower level of education. The results emphasise the growing importance of AI skills in influencing hiring decisions.