AI literacy and deepening thinking

By Allen Yang

November 14, 2025

About this collection

This collection examines the transformative impact of AI and GenAI on knowledge work, research, and education, revealing both opportunities and challenges. **Core themes** include productivity gains (12-40% improvements in consultant performance), capability expansion beyond current skill levels, and the emergence of new human-AI collaboration patterns ('Centaurs' vs 'Cyborgs'). However, the collection highlights critical tensions: AI struggles with tasks outside its capabilities (19% lower correctness), creates new digital divides particularly affecting marginalized communities, and threatens research integrity through authorship ambiguity and potential fabrication. **The literacy gap emerges as crucial**—fear (52% nervous) rivals excitement (54%) about AI, with underrepresented groups facing disproportionate barriers. The collection suggests AI literacy isn't just technical knowledge but encompasses ethical considerations, responsible use, and an 'engineering mindset' for effective supervision. **Strategic implications**: Organizations must balance productivity gains against over-reliance risks, while educational systems need comprehensive frameworks (like AILit's 22 competences) to prepare workers and learners. The research automation tools (SPARK) demonstrate AI's potential to transform systematic review processes, yet underscore the need for human oversight in maintaining scholarly integrity.

Curated Sources

AI-pocalypse now: Automating the systematic literature review with SPARK (Systematic processing and automated review Kit) – gathering, organising, filtering, and scaffolding. - ScienceDirect

Researchers face challenges due to the exponential growth of global research outputs and the advent of Generative Artificial Intelligence (GenAI). Traditional literature review methods are becoming outdated, and new tools are needed to automate the process. This article introduces SPARK (Systematic Processing and Automated Review Kit), a suite of computational-based approaches designed to automate the collection, organisation, and filtering of journal articles for systematic literature reviews. SPARK utilises Python scripts to extract data from Web of Science, Scopus, and Google Scholar, and applies Latent Dirichlet Allocation (LDA) topic modelling to identify hidden themes in article abstracts. The methodology is demonstrated through a case study on trauma-informed policing, resulting in the automation of the front-end stages of a systematic literature review.

Key Takeaways

SPARK automates the initial stages of systematic literature reviews, enhancing research efficiency.
The tool uses hard-coded scripts to maintain control and accuracy, avoiding GenAI-induced errors.
LDA topic modelling identifies hidden themes in article abstracts, facilitating data extraction template construction.
The methodology is adaptable and can be customised for various research needs and standards.
SPARK promotes transparency, accessibility, and collaboration in research through open science principles.

GenAI Increases Productivity & Expands Capabilities | BCG

Generative AI (GenAI) is not just a productivity tool but can expand workers' capabilities beyond their current skill levels. A recent experiment by the BCG Henderson Institute tested nontechnical knowledge workers' ability to perform data-science tasks with GenAI assistance. The study involved 480 BCG consultants who were given access to Enterprise ChatGPT-4 with Advanced Data Analysis Feature or traditional resources. Results showed that GenAI-augmented workers could instantly expand their aptitude for new tasks, achieving significant performance improvements in coding, statistical understanding, and predictive modeling. Participants with moderate coding experience performed better across all tasks, even non-coding ones, suggesting an 'engineering mindset' is crucial. The study highlights implications for talent acquisition, internal mobility, employee learning, teaming, and performance management. Leaders must manage risks associated with over-reliance on GenAI and ensure workers have necessary background knowledge to supervise AI outputs. The findings suggest a future where GenAI enables workers to take on complex tasks beyond their current capabilities, requiring organizations to rethink workforce planning and skill development strategies.

Key Takeaways

GenAI can instantly expand workers' capabilities for complex tasks beyond their current skill levels, achieving significant performance improvements.
Moderate coding experience is a key success factor for workers using GenAI, even for non-coding tasks, suggesting an 'engineering mindset' is crucial.
Organizations must manage risks associated with GenAI, including over-reliance and lack of background knowledge to supervise AI outputs.
The study's findings have significant implications for talent acquisition, internal mobility, employee learning, teaming, and performance management in a GenAI-augmented workforce.

The Impact of Generative Artificial Intelligence on Research Integrity in Scholarly Publishing - The American Journal of Pathology

Generative artificial intelligence (Gen AI) and large language models (LLMs) have significantly impacted scholarly publishing and research integrity since the launch of ChatGPT in November 2022. Gen AI poses challenges to research integrity, including authorship issues, plagiarism, data fabrication, and image manipulation. While Gen AI can facilitate cheating, it also offers opportunities to improve research integrity through AI-based detection tools and enhanced editorial processes. The responsible use of Gen AI, together with full disclosure, is crucial to maintaining trust and transparency in scholarly publishing.

Key Takeaways

Gen AI challenges traditional notions of authorship and accountability in research.
AI-based detection tools can help identify and prevent research misconduct.
Responsible use of Gen AI is essential to maintaining research integrity.
Gen AI can both facilitate and combat research misconduct.
The future of scholarly publishing requires balancing innovation with integrity.

Empowering Learners for the Age of AI

The AI Literacy Framework (AILit Framework) is a joint initiative of the European Commission and the Organization for Economic Cooperation and Development (OECD), supported by Code.org and international experts. It aims to empower learners for the age of AI by providing a comprehensive framework for AI literacy in primary and secondary education. The framework defines AI literacy as the technical knowledge, durable skills, and future-ready attitudes required to thrive in a world influenced by AI. It encompasses four domains: Engaging with AI, Creating with AI, Managing AI, and Designing AI, which together outline 22 competences for learners. The framework emphasizes understanding AI's technical foundations, human skills for collaboration with AI, and ethical considerations. It is designed to be interdisciplinary, global, foundational, practical, illustrative, and durable, preparing learners to responsibly interact with AI and navigate its societal impacts.

Key Takeaways

The AILit Framework emphasizes the importance of understanding AI's technical foundations, including its reliance on data and statistical inferences, to critically evaluate its capabilities and limitations.
It highlights the need for human skills such as critical thinking, creativity, and computational thinking to collaborate effectively with AI tools.
The framework stresses ethical considerations, including recognizing AI's potential to replicate harmful biases and its environmental impact, to ensure responsible AI use.
The four domains of AI literacy - Engaging with AI, Creating with AI, Managing AI, and Designing AI - provide a comprehensive approach to developing learners' AI competences.
Educators play a crucial role in integrating AI literacy into their teaching practices, requiring targeted support to build their own AI competences and develop effective pedagogies.

REVIEW ARTICLE

Artificial intelligence (AI) is transforming various fields, making AI literacy crucial for learners. This study reviews AI literacy education research from 2014 to 2024 using bibliometric analysis. The research shifted from an exploratory phase to rapid growth, with four distinct developmental trajectories emerging. Nine prominent research themes were identified, including data literacy, machine learning, and AI literacy. The study highlights the interdisciplinary nature of AI literacy education and its connections to information, digital, and algorithmic literacy. Key findings include the exponential growth of publications, the importance of integrating AI ethics into education, and the need for collaborative approaches to develop comprehensive AI literacy frameworks.

Key Takeaways

AI literacy education research has experienced exponential growth since 2018, indicating increasing academic interest and impact.
Four developmental trajectories in AI literacy research have emerged, emphasizing interdisciplinary connections to information, digital, and algorithmic literacy.
Nine prominent research themes have been identified, with data literacy, machine learning, and AI literacy being focal points, highlighting AI's evolving role in education.
The study underscores the need for integrating AI ethics into educational frameworks to enhance AI literacy and promote responsible AI use.
Future research should prioritize assessing and promoting AI literacy among diverse age groups, including adults in specialized sectors.

AI literacy and the new Digital Divide - A Global Call for Action | Global AI Ethics and Governance Observatory

The rapid advancements in artificial intelligence (AI) have widened the digital divide, creating an AI divide that disproportionately affects marginalized communities. To bridge this gap, promoting AI literacy is crucial. Global leaders must spearhead efforts to develop and implement local educational programs that teach AI basics, decrease fear, and increase curiosity. Key actions include resource allocation to trusted nonprofits and educational institutions, local engagement through community-driven initiatives, inclusive education materials, collaborative efforts between governments and tech companies, and promoting continuous learning. The goal is to equip individuals with essential AI skills to thrive in an AI-driven world.

Key Takeaways

The AI divide exacerbates existing social inequalities, particularly affecting women, people of color, and other marginalized groups who face unequal access to AI benefits and opportunities.
Fear and lack of understanding are significant barriers to AI adoption, with nearly equal numbers of adults reporting being nervous (52%) and excited (54%) about AI products and services.
Targeted AI literacy programs are essential to support vulnerable groups and address the skills gap, particularly in the workforce where women are more likely to be exposed to AI-related job changes.
Global leaders must promote AI literacy through local educational programs, resource allocation, and community engagement to create a foundation of understanding and decrease fear.
Collaborative efforts between governments, tech companies, and educational institutions are critical to amplify the reach and impact of AI literacy programs.

Working Paper 24-013

The study examines AI's impact on knowledge-intensive tasks through a field experiment with 758 consultants. AI access significantly increased productivity (12.2% more tasks completed, 25.1% faster) and quality (40% higher). Consultants below average performance threshold benefited most (43% increase). AI struggled with tasks outside its capabilities, leading to 19% lower correctness. Two human-AI collaboration patterns emerged: 'Centaurs' dividing tasks between human and AI, and 'Cyborgs' integrating AI into their workflow.

Key Takeaways

AI significantly boosts productivity and quality within its capability frontier
Consultants below average performance threshold benefit most from AI assistance
AI can decrease performance when used outside its capability frontier
Two distinct human-AI collaboration patterns emerge: Centaurs and Cyborgs
Centaurs divide tasks between human and AI, while Cyborgs integrate AI into their workflow

Frequently Asked Questions

How does the 'engineering mindset' identified in the BCG consultant study relate to the AILit Framework's 22 competences—specifically, which competences from 'Managing AI' and 'Designing AI' domains correlate with the coding experience advantage?
What explains the tension between AI expanding capabilities beyond current skill levels (BCG study) while simultaneously creating a 19% failure rate on out-of-scope tasks—and how should organizations calibrate task allocation given this dual reality?
The BCG study shows below-average performers gained 43% improvement versus 12.2% overall—how does this differential impact interact with UNESCO's finding that marginalized communities face the greatest AI literacy barriers? Does AI risk widening or narrowing performance gaps?
How do the 'Centaur' versus 'Cyborg' collaboration patterns map onto the four AILit Framework domains (Engaging, Creating, Managing, Designing)—and which pattern requires more advanced competences?
SPARK automates systematic literature review front-end stages using LDA topic modeling—could similar computational approaches address the research integrity challenges identified in scholarly publishing, or would automation exacerbate fabrication and plagiarism risks?
UNESCO emphasizes 'trusted local sources' for AI literacy over 'impersonal online resources'—how does this community-driven approach reconcile with the need for standardized competences in the AILit Framework's global, foundational design?
The BCG study found GenAI-augmented workers achieved 40% higher quality—but how should 'quality' be measured when the American Journal of Pathology warns about AI-generated authorship ambiguity and data fabrication in research outputs?
What's the relationship between the 'supervision' skills needed to manage AI's 19% failure rate and the AILit Framework's 'Managing AI' domain—specifically, which of the 22 competences address detecting when AI operates outside its capabilities?
How do the two human-AI collaboration patterns ('Centaurs' dividing tasks vs 'Cyborgs' integrating workflow) perform differently on tasks outside AI's capabilities, and what does this suggest about risk management strategies?
UNESCO identifies fear (52% nervous) as a barrier to AI adoption, while the BCG study shows significant productivity gains—is there evidence that fear correlates with more cautious, effective AI use that avoids the 19% failure rate on out-of-scope tasks?