Tasks Versus Skills Part 3: Let Learning Breathe
A 7-part series on AI, the state of tasks versus skills, and some provocative what-if concepts for Learning and Development, HR and Talent teams.
This is Part 3 of a series called Tasks Versus Skills: Part 1: Introduction, Part 2: Playbook, Part 3: Let Learning Breathe, Part 4: Task Intelligence Control Room, Part 5: Tasks as EX Product, Part 6: IQA Prototype, Part 7: Talent Is Not a Commodity; Google Books “Tasks vs Skills” free ebook (Parts 1-6 only).
What-If: Let Learning Breathe
Liberating content from the container of duration, decay, and bulk proficiency
(Photo by Guillaume Bolduc on Unsplash)
Maya’s reverence for wine, heritage, sunshine, and vigor resembles, for me, the respect we all have for a well-produced, bottled, and enriching learning experience. Many of us in L&D are similar connoisseurs of fine content, just as Maya is for a fine Syrah or Miles is for his 1961 Château Cheval Blanc St. Emilion Grand Cru, which he keeps locked up in a closet until a special occasion.
What seems relatable is how the quality and experience of a training course, any course, may have parallel characteristics with those of a bottle of wine once uncorked and poured for learners.
In a way, whether it’s a dense leadership course provoking constructive debate or a lighter five-minute product update training video, the course does not become activated or alive until it’s exposed and experienced by people. Moreover, to Maya’s point, how you’re exposed to a learning experience one day will “taste different than if I opened up any other day”.
Yet, similar to Miles’ 1961 Château Cheval Blanc, most training courses have been in a hold, in containers of fixed time or duration and fixed scope. It’s a 4-hour online course. It’s only for engineers. It specifically addresses XYZ generic scenarios or use cases. Its design may inherently contain a bias to ensure clarity with clear instructions, yet with limited objectives and outcomes. When considering skills, especially constantly evolving and perishable skills, residing in a container of restrained utility can lead to a limited or stale impact.
As shared, when considering large language models (LLMs), retrieval augmented generation (RAG), Agentic AI, and their appeal for task-centric instruction, will we soon have new methods to access knowledge across the company, perhaps in real-time. Can we train AI agents, probably hundreds of task-specific agents, to consistently harvest fresh insights and reassemble them into new instructional and informational assets?
Would this change L&D approaches and learners’ experience to be less time-bound by minutes and hours, “constantly evolving” and harvesting relevant insights from teammates, and dynamically reassembling with highest priority skills?
Let’s step back for a minute and understand how we got here, containers and otherwise.
Suppose we work off the premise that workforce development or corporate training exists to improve productivity, develop skills and knowledge to stay competitive and grow careers, foster innovation, gain efficiencies, create new leaders, and mitigate risks. What does that mean to the training programs we provide?
Let’s consider four critical factors that influenced organizational goals for the twentieth century, as defined by emerging themes in the seminal book The Boundaryless Organization. The influencing business themes:
Company Size: As a company grows, it has a greater ability to produce more, leverage capital, and stay competitive.
Role Clarity: Ensure clear distinctions with work to be conducted, roles and levels of access and authority are clear, and job expectations were designed to gain efficiencies with the tasks and proficiency required.
Specialization: Work, individuals, teams, and functions such as Finance or Operations are finely tuned, enhancing their disciplines.
Control: With these three parameters ideally in place, companies need a set of controls to ensure consistency in work and predictability in scope, speed, and target attainment.
Many suggest that these “controlling” business factors have led to training programs with similar intentions or directions. Perhaps these intentions have led to insufficient or perishable programs not meeting the dynamic needs of modern work and workers. In many or most cases, this lack may be driven by a lack of understanding of fundamental learning design principles, too much focus on content volume versus context and quality, and of course external business drivers such as macroeconomics etc.
As learning leader Clark Quinn shares in Revolutionize Learning & Development, with candor: “Let me be blunt, the current learning and development (L&D) industry is failing. Badly. Overall, L&D is only doing a fraction of what it could and should be doing, and the part that it is doing, it is doing poorly…. Yet the potential exists, particularly in this emerging state of change…”
The way we work is shifting, and given the advent of AI, we need to move away from these controlling and containing factors to smarter models, still addressing C-level expectations for L&D to show ROI, as well as innovative approaches we L&D-ers are anxious to explore.
A from-to review of where I see us potentially heading:
So, how would this work?
What I propose is a new, somewhat alternative world where a set timeframe and limited structure do not restrict learning relevance and related courseware. Nor will it be controlled for you. It would not be stagnant or prone to becoming stale. It is not encased to ensure repeatability, uniformity, and scale. Given the speed and acceleration of business, competition, and skills, content would not be destined to perish.
It would constantly harvest, contextualize, and regenerate new assets based on knowledge and data gathered across the enterprise. This ideal state would provision learning content and assets immediately, instantaneously, and concurrently with your work activities. This is, in essence, a new paradigm where learning content is alive, breathing, and never restrained in old-school containers.
This would represent an extension of Lifelong or Continuous Learning where learning new knowledge and skills is certainly ongoing and may be tied to a Growth Mindset culture. Yet, the content itself would have these new AI-provisioned attributes:
Instead of a limited, named group of experts who create a course, AI technologies would dynamically gather content from multiple sources within the enterprise, including experts’ documents and assets, and ‘best’ practices yet discovered. E.G., gathered from reports and spreadsheets, emails, portals, Sharepoint sites, social channels such as Teams or Slack, and monitored systems. If AI notetakers are recording meeting transcripts and directions, this could be harmonized for learners.
Instead of solely consuming content from a learning platform or learning management system, the informational and instructional assets are available within one’s ‘in-the-flow-of’ work or self-generated based on an individual’s need.
Rather than receiving one-size-fits-all offerings, task-specific AI agents tailor and personalize instruction based on role, use case, and context.
Finally, your learning offering is dynamically monitored, categorized, reassembled, and readily available, supporting one’s ever-changing key work requirements or most critical Moments of Need:
(Image from 5MomentsofNeed)
As shared previously, autonomous AI agents and Retrieval Augmented Generation or RAG methods will access a company’s infrastructure, including for example, Office 365 offerings, knowledge databases, Human Resources, Sales, Customer, and Manufacturing systems. Is this personally invasive? Maybe. Reminder: Your company owns workers’ intellectual property and related data in almost all workplace scenarios. This is a reality that, frankly, we should take advantage of.
Here are three AI-enabling ‘task methods’ that may help this first what-if concept breathe and come to life [pun intended]. All methods identify desired tasks, connect via tasks, assemble by tasks, and fuel related skills.
Task Method 1. LLM Knowledge Graphs for Learning
An LLM knowledge graph is a structured representation of information where concepts, entities, and their relationships are organized in an interconnected network. In the context of training content, an LLM knowledge graph can identify, harvest, and reassemble disparate pieces of content into diverse forms of training.
(Image: Towardsdatascience)
How would a knowledge graph operate and ensure relevant content (though dispersed and unstructured) is harvested, reassembled, and provided to the learner, ideally in the moment of need? Six potential steps:
Learning requirements are determined. To fuel the graph, we need to be clear about our goals and how to set proper guidelines. This can be in the form of instructions or more detailed hard prompts.
Data Ingestion and Integration: Data is identified and ingested from multiple enterprise sources, including emails, documents, presentations, messaging apps, and existing training materials, into an LLM knowledge graph framework and system.
Entity and relationship extraction: The LLM analyzes the ingested data, identifying key entities, topics, and their relationships. It creates nodes in the knowledge graph that links related information across the various content types.
Dynamic content & task mapping: As these instructional nodes source relevant data and to-do tasks, new mappings are developed (“Identify and assemble all Product Marketing insights from recent meetings and documents related to product XYZ’s value proposition, and prepare a production outline for a Colossyan sales training video…”).
Task-instructed context harvesting: As dozens of similar learning requests and instructions are collected, new nodes reshape the graph, enabling real-time tracking of emerging concepts and dynamically linking them to existing content. AI agents will be tasked to leverage the most a) high-valued content from b) the highest-ranked nodes (i.e., experienced Marketers’ activities c) by querying the graph for targeted topics and related tasks and skills. This connects disparate pieces of content into logical, contextually aligned clusters.
Reassemble into adaptive learning materials: Newly tagged and harvested content will supply learning modules that evolve with new data, providing adaptive, up-to-date, and contextually enriched training content that reflects the organization's living knowledge.
There is related momentum to track. For example, Microsoft’s use of similar graphs is being incorporated into its 365 learning tools via its Graph API. IBM is also expanding its Knowledge Catalog using similar approaches across the company. This is advancing, especially with Task Method 2: DocETL Framework and Unstructured-to-Structured RAG-to-LLM.
(https://learn.microsoft.com/en-us/graph/overview)
Task Method 2. DocETL Framework and Unstructured-to-Structured RAG-to-LLM
ETL (Extract, Transform, Load) is a data integration process that extracts data from various sources, transforms it into a desired format or structure, and loads it into a target system for analysis or operational use.
DocETL Framework
First, if we consider LLM knowledge graphs our strategic approach, how do we test, trial, and ideally operationalize them?
The DocETL framework (ETL meaning Extract, Transform, Load) is a method created by the EPIC Data Lab at UC Berkeley, and it is an innovative approach designed for complex document processing. DocETL is intended to optimize and improve the accuracy of analyzing unstructured data, including diverse content such as reports and text files, and leverage your company’s LLM of choice. It has the potential to enable this vision of a learning environment where knowledge is continuously harvested, contextualized, and regenerated, breaking down complex tasks into manageable steps using automated agents.
DocETL makes it easier for organizations to turn unstructured data (again, documents, emails, portals, social channels, and AI notetaker’s meeting notes…) into meaningful insights. It ensures that content is always fresh, personalized, and aligned with real-time business needs, supporting critical moments of need and enhancing workforce learning.
In this example, DocETL employs LLMs to synthesize diverse reports, map similarities, and generate a consolidated and context-sensitive result. DocETL is also open-sourced and can be applied across various domains, from legal documents to help desk guides, workflow processes, and more.
(Image: https://epic.berkeley.edu/publications)
How DocETL may support our vision for an ‘uncontained’ learning environment:
Adaptive learning content creation: Using DocETL to process content from multiple enterprise sources continuously, L&D teams can automatically generate learning materials that are current, relevant, and specific to the employees' real-time needs.
Contextual learning opportunities: DocETL’s ability to decompose complex tasks and documents could allow L&D to create granular, contextualized learning opportunities. For instance, information could be tailored to an employee's specific workflow within the enterprise, enhancing relevance and supporting "in-the-flow-of-work" learning.
Efficiency and scalability in content development: DocETL's optimization features can streamline learning content development by eliminating manual intervention in content curation. It identifies, synthesizes, and validates sourced content, reducing the burden on content creators and ensuring the material is accurate and comprehensive.
Personalization at Scale: DocETL's ability to dynamically reconfigure and evaluate data processing pipelines allows L&D teams to create personalized learning paths for different roles, tasks, or use cases, adjusting the learning materials in response to the individual learner's data and evolving skill needs.
Unstructured-to-Structured RAG-to-LLM
Important recap…. RAG (Retrieval-Augmented Generation) is an AI approach that combines pulling relevant information from a database or documents (retrieval) responses are accurate and grounded in real data. A Large Language Model (LLM) is an AI system trained on vast amounts of text to understand and generate human-like language for tasks like answering questions, writing, or summarizing information.
So, what could this mean for our Let Learning Breathe What-If?
The work being conducted by Unstructured Technologies, one of Fast Company’s 2024 Most Innovative Companies is truly remarkable–similar to DocETL technologies. Unstructured Technologies has developed an open-source platform that converts complex, unstructured data formats—such as PDFs and company documents—into AI-friendly JSON files. JSON (JavaScript Object Notation) is a fairly easy-to-read format used to organize and exchange data. By automating the transformation of raw data into structured formats, Unstructured streamlines the assembly of unstructured content into coherent training courses and instructional materials, enhancing efficiency and effectiveness in content development.
The premise again is to optimize the analysis of unstructured data. This includes diverse content from enterprise databases, files, videos, emails and meeting transcripts such as reports and text files, and restructure into assets powered by your company’s LLM of choice.
Unstructured’s approach is somewhat identical and follows these general steps:
Unstructured's solution aligns seamlessly with our Let Learning Breathe thesis by enabling dynamic transformation of unstructured data into structured, AI-ready formats. This capability supports a paradigm where learning content is not static but continuously harvested, contextualized, and regenerated. This ensures that instructional assets remain alive, relevant, and adaptable to the rapid pace of business and evolving workforce needs, fostering a truly continuous and responsive learning ecosystem.
Task Method 3. The NEAR FUTURE: Graph-Based Generative AI
(Image: buildprompt.ai)
“By blending generative AI with graph-based computational tools, this approach reveals entirely new ideas, concepts, and designs previously unimaginable. We can accelerate scientific discovery by teaching generative AI to make novel predictions about never-before-seen ideas, concepts, and designs.” Markus Buehler.
Professor Buehler’s MIT research “Accelerating scientific discovery with generative knowledge extraction, graph-based representation, and multimodal intelligent graph reasoning” reveals a method describing how diverse forms of information may have strong, undiscovered connections, with groups of interrelated themes, contexts, and insights. This advanced AI method is based on category theory, which is a mathematical approach that examines the relationships of disconnected, abstract concepts and creates “unifying diverse systems” frameworks. With task-based instruction, this model can analyze 1000s of assets and create a new knowledge map that can assemble what appears to be disjointed information into valuable instruction. “With this advanced AI model, scientists can draw insights from music, art, and technology to analyze data from these fields to identify hidden patterns that could spark a world of innovative possibilities for material design, research, and even music or visual art.” [27]
A recent example of graph-based GenAI is creating a new biological substance based on abstract patterns in a Wassil Kandinsky painting, “Composition VII.” Here, Buehler’s graph-based model built a new mycelium-based composite that may lead to new biodegradable technologies, building methods, and biomedical devices. All of this stems from and is inspired by an impressionist artist’s abstract work:
(Image: Wassily Kandinsky (left), Markus Buehler, with the assistance of his new artificial intelligence system [center and right])
Buehler adds: “Graph-based generative AI achieves a far higher degree of novelty, explorative of capacity and technical detail than conventional approaches, and establishes a widely useful framework for innovation by revealing hidden connections… This study contributes to the field of bio-inspired materials and mechanics. It sets the stage for a future where interdisciplinary research powered by AI and knowledge graphs may become a scientific and philosophical inquiry tool as we look to other future work.”
How does this align with our Let Learning Breathe vision? It may appear a bit early to identify ties to gluing disparate content, providing task-instructed reassembly for instructional needs. This approach could lead to more innovative, personalized, and adaptive learning experiences that break free from traditional constraints and continuously evolve with the organization's needs.
Graph-based generative AI examples affecting Learning & Development:
Continuous and diverse knowledge harvesting and regeneration: The graph-based AI allows complex relationships to be visualized, understood, and leveraged in real-time—aligning with a "breathing" learning ecosystem. Further, this model can represent and discover knowledge across diverse domains.
Cross-disciplinary insights for learning: Buehler’s work illustrates how AI can connect seemingly unrelated content, work, and people domains. This aligns with our thesis that learning content could link diverse sources like business processes, technical and forecasting etc.
Question traditional learning boundaries: Traditional learning, often constrained by uniformity, static content, and fixed structure, seems eager and willing to break down these boundaries. As Buehler continues: “Researchers can use this framework to answer complex questions, find gaps in current knowledge, suggest new materials designs, predict how materials might behave, and link concepts that had never been connected before.”
For curious and investigative L&D teams, our first task-centric What-If concept, Let Learning Breathe, will have many benefits. Explore! These benefits or values exceed basic or traditional efficiencies and effectiveness protocols. This again includes broader impacts on cost controls, operations and operational efficiencies, resource allocation, and less reliance on external Buy models. Most apparent OPPORTUNITIES for Learning & Development from What-If #1, overall:
Automated content curation
Improved content relevance
Scalability
Real-Time content updates
Enhanced analytics
Broader, horizontal (and ideally seamless) integration
Operational efficiencies and improved training standards and governance
Lastly, an important reminder that this approach is not about making text the be-all and end-all solution. As shared, LLMs can create multimodal or multiple formats of entertainment, education, and generally creative stuff. In a very idealistic state, I can envision autonomous AI agents constantly looking for prescribed insights that can be dynamically reassembled in near real-time.
(Image: spotintelligence)
However, this dynamic, breathing AI-driven concept, at this point, remains a concept, though, for me, it is worth an optimistic wager. That said, currently, we do need to stay cognizant of the challenges a Let Learning Breathe approach may contain:
Potential for information overload
Risk of irrelevance
Loud, in your face Continuous Learning
Inequality of access or use
Human-conducted quality controls
Less human-to-human collaboration
Invalid sources or source accuracy
Privacy and confidential information
“Human projection” or judging GenAI from a human lens; great Harvard report: Human Learning about AI
I hope this first What-If may remove some boundaries in accessible learning insights and helpful practices. This could be a game-changer for many of us and for the business, goals, and teammates we enable. This model ensures that learning can keep pace with the speed of change, our competition, and skill demands and eventually support every Moment of Need.
Moreover, we have an obligation to remove the limitations or containers that limit true, individual growth.
Our second What-If concept goes deeper into our ability to fix, forecast, and, if possible, predict business outcomes–both good and not-so-good–via a more robust task Command Center approach to tasks and skills.
Links: Part 1: Introduction, Part 2: Playbook, Part 3: Let Learning Breathe, Part 4: Task Intelligence Control Room, Part 5: Tasks as EX Product, Part 6: IQA Prototype, Part 7: Talent Is Not a Commodity; Google Books “Tasks vs Skills” free ebook (Parts 1-6 only).
Special thanks to colleagues who guided me as contributors and reviewers, and to so many who have inspired my thinking and curiosity on this subject.
Contributors: Cathy Moore, Clark Quinn, Felipe Hessel, Gianni Giacomelli, Giri Coneti, Jon Fletcher, Julie Dirksen, Megan Torrance, Nick Shackleton-Jones, Nina Bressler, Will Thalheimer, Ross Dawson
Inspirations: Allie K. Miller, Amanda Nolen, Andrew Kable (MAHRI), Bhaskar Deka, Brandon Carson, Brian Murphy, Chara Balasubrmanian, Dani Johnson, Darren Galvin, Dave Buglass Chartered FCIPD, MBA, Dave Ulrich, David Green 🇺🇦, David Wilson, Deborah Quazzo, Dennis Yang, Detlef Hold, Donald Clark, Donald H Taylor, Dr Markus Bernhardt, Dushyant Pandey, Egle Vinauskaite, Emma Mercer (Assoc CIPD, MLPI), Ethan Mollick, Gordon Trujillo, Guy Dickinson, Harish Pillay, Hitesh Dholakia, Isabelle Bichler-Eliasaf, Isabelle Hau, Joel Hellermark, Joel Podolny, Johann Laville, Jon Lexa, Josh Bersin, Josh Cavalier, Joshua Wöhle, Julian Stodd, Karen Clay, Karie Willyerd, Kate Graham, Kathi Enderes, Marga Biller, Marc Zao-Sanders, Meredith Wellard, Mikaël Wornoo🐺, Nico Orie, Noah G. Rabinowitz, Nuno Gonçalves, Oliver Hauser, Orsolya Hein, Patrick Hull, Peter Meerman, Peter Sheppard, Dr Philippa Hardman, Raffaella Sadun, Ravin Jesuthasan, CFA, FRSA, René Gessenich, Ross Dawson, Ross Garner, Sandra Loughlin, PhD, Simon Brown, Stacia Sherman Garr, Stefaan van Hooydonk, Stella Collins, Trish Uhl, PMP 👋🏻, Tony Seale, Zara Zaman
Acknowledging leveraging Perplexity, Gemini, ChatGPT, and Claude for research, formatting, testing links, challenging assumptions and aiding the creative process.
LICENSING
Unless otherwise noted, the contents of this series are licensed under the Creative Commons Attribution 4.0 International license.
Should you choose to exercise any of the 5R permissions granted you under the Creative Commons Attribution 4.0 license, please attribute me in accordance with CC's best practices for attribution.
If you would like to attribute me differently or use my work under different terms, contact me at https://www.linkedin.com/in/marcsramos/.
NOTES
Part 5: Let Learning Breathe
[27] reveals a method describing how diverse forms of information: Markus J Buehler, (2024), “Accelerating scientific discovery with generative knowledge extraction, graph-based representation, and multimodal intelligent graph reasoning”, Mach. Learn.: Sci. Technol. 5 035083, https://iopscience.iop.org/article/10.1088/2632-2153/ad7228
ADDITIONAL RESOURCES (ALL 7 PARTS OF THIS SERIES)
Wen, J., Zhong, R., Ke, P., Shao, Z., Wang, H., & Huang, M., (2024), “Learning task decomposition to assist humans in competitive programming”, Proceedings of the 2024 Conference of the Association for Computational Linguistics (ACL 2024), https://arxiv.org/pdf/2406.04604
Wang, Z., Zhao, S., Wang, Y., Huang, H., Shi, J., Xie, S., Wang, Z., Zhang, Y., Li, H., & Yan, J., (2024), “Re-TASK: Revisiting LLM tasks from capability, skill, and knowledge perspectives”, arXiv, https://arxiv.org/abs/2408.06904
Hubauer, T., Lamparter, S., Haase, P., & Herzig, D., (2018), “Use cases of the industrial knowledge graph at Siemens”, Proceedings of the International Workshop on the Semantic Web, CEUR Workshop Proceedings, https://ceur-ws.org/Vol-2180/paper-86.pdf
Erik Brynjolfsson, Danielle Li, Lindsey R. Raymond, “Generative AI At Work”, http://www.nber.org/papers/w31161 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 April 2023, revised November 2023, https://www.nber.org/system/files/working_papers/w31161/w31161.pdf
https://www.royermaddoxherronadvisors.com/blog/2016/april/the-knowledge-worker-in-healthcare/
Connie J.G Gersick, J.Richard Hackman, (1990), “Habitual routines in task-performing groups, https://www.sciencedirect.com/science/article/pii/074959789090047D
Organizational Behavior and Human Decision Processes”, Volume 47, https://doi.org/10.1016/0749-5978(90)90047-D
World Economic Forum, (2023), “Future of Jobs Report 2023”, World Economic Forum, https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf
Boston Consulting Group, (2024), “How people can create—and destroy—value with generative AI”, BCG, https://www.bcg.com/publications/2023/how-people-create-and-destroy-value-with-gen-ai
Kaisa Savola, (2024), Challenges Preventing Organizations from Adopting a Skills-Based Model”, LinkedIn, https://www.linkedin.com/pulse/challenges-preventing-organizations-from-adopting-savola-she-her--24c6f/
Patel, A., Hofmarcher, M., Leoveanu-Condrei, C., Dinu, M.-C., Callison-Burch, C., & Hochreiter, S., (2024), “Large language models can self-improve at web agent tasks”, arXiv, https://arxiv.org/abs/2405.20309
Sagar Goel and Orsolya Kovács-Ondrejkovic, (2023), “Your Strategy Is Only as Good as Your Skills”, BCG Publications, https://www.bcg.com/publications/2023/your-strategy-is-only-as-good-as-your-skills
“Workplace 2.0: The Promise of the Skills-Based Organization”, (2024), Udemy Business, https://info.udemy.com/rs/273-CKQ-053/images/udemy-business-workplace-2.0-promise-skills-based-organization.pdf?version=0
Shakked Noy and Whitney Zhang (2023): “Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence”, MIT Economics Department working paper. Retrieved March 13, 2023 from https://economics.mit.edu/sites/default/files/inline-files/Noy_Zhang_1_0.pdf
“AI at Work Is Here. Now Comes the Hard Part.”, (2024), 2024 Work Trend Index Annual Report from Microsoft and LinkedIn, https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part
“LinkedIn Workplace Learning Report”, (2024), LinkedIn Learning, https://learning.linkedin.com/content/dam/me/business/en-us/amp/learning-solutions/images/wlr-2024/LinkedIn-Workplace-Learning-Report-2024.pdf
Macnamara BN, Berber I, Çavuşoğlu MC, Krupinski EA, Nallapareddy N, Nelson NE, Smith PJ, Wilson-Delfosse AL, Ray S. “Does using artificial intelligence assistance accelerate skill decay and hinder skill development without performers' awareness?”, Cogn Res Princ Implic. 2024 Jul 12;9(1):46. doi: 10.1186/s41235-024-00572-8. PMID: 38992285; PMCID: PMC11239631, https://pmc.ncbi.nlm.nih.gov/articles/PMC11239631/
Matt Beane, (2024), “The Skill Code: How to Save Human Ability in an Age of Intelligent Machine”, Harper, https://mitsloan.mit.edu/ideas-made-to-matter/how-can-we-preserve-human-ability-age-machines
Lazaric, N., (2012), “Routinization of Learning”, Encyclopedia of the Sciences of Learning. Springer, Boston, MA. https://doi.org/10.1007/978-1-4419-1428-6_246
Hering, A., & Rojas, A., (2024), “AI at Work: Why GenAI Is More Likely To Support Workers Than Replace Them.”, Indeed Hiring Lab, https://www.hiringlab.org/2024/09/25/artificial-intelligence-skills-at-work/
Morgan, T. P., (2024), “How GenAI Will Impact Jobs In the Real World”, HPCwire, https://www.hpcwire.com/2024/09/30/how-genai-will-impact-jobs-in-the-real-world/
https://trainingindustry.com/magazine/fall-2022/improving-the-productivity-of-knowledge-workers/
Balasubramanian, R., Kochan, T. A., Levy, F., Manyika, J., & Reeves, M., (2023), “How People Can Create—and Destroy—Value with Generative AI”, Boston Consulting Group, https://www.bcg.com/publications/2023/how-people-create-and-destroy-value-with-gen-ai
Marcin Kapuściński, (2024), “Boosting productivity: Using AI to automate routine business tasks”, TTMS, https://ttms.com/boosting-productivity-using-ai-to-automate-routine-business-tasks/
https://scheibehenne.com/ScheibehenneGreifenederTodd2010.pdf
Paulo Carvalho, (2024), “How Generative AI Will Transform Strategic Foresight”, IF Insight & Foresight, https://www.ifforesight.com/post/how-generative-ai-will-transform-strategic-foresight
Deok-Hwa Kim, Gyeong-Moon Park, Yong-Ho Yoo, Si-Jung Ryu, In-Bae Jeong, Jong-Hwan Kim, (2017), “Realization of task intelligence for service robots in an unstructured environment”, Annual Reviews in Control, Volume 44, ISSN 1367-5788, https://doi.org/10.1016/j.arcontrol.2017.09.013.
Kweilin Ellingrud, Saurabh Sanghvi, Gurneet Singh Dandona, Anu Madgavkar, Michael Chui, Olivia White, Paige Hasebe, (2023), “Generative AI and the future of work in America”, McKinsey Global Institute, https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america#/
MIT Technology Review Insights, (2024), “Multimodal: AI’s new frontier”, https://wp.technologyreview.com/wp-content/uploads/2024/05/Jina-AI-e-Brief-v4.pdf?utm_source=report_page&utm_medium=all_platforms&utm_campaign=insights_ebrief&utm_term=05.02.24&utm_content=insights.report
Sandra Ohly, Anja S. Göritz, Antje Schmitt, (2017), “The power of routinized task behavior for energy at work”, Journal of Vocational Behavior, Volume 103, Part B, https://doi.org/10.1016/j.jvb.2017.08.008
Baddeley, A.D. (2003). Working memory: looking back and looking forward. Nature Reviews Neuroscience, 4, p.829-839, https://pubmed.ncbi.nlm.nih.gov/14523382/
Baddeley, A.D. and Hitch, G. (1974). Working Memory. Psychology of Learning and Motivation, 8, p.47-89, https://www.sciencedirect.com/science/article/pii/S0960982209021332
Chandler, P. and Sweller, J. (1991). Cognitive Load Theory and the Format of Instruction. Cognition and Instruction, 8 (4), p. 293-332, https://www.tandfonline.com/doi/abs/10.1207/s1532690xci0804_2
Chandler, P. and Sweller, J. (1992). The split-attention effect as a factor in the design of instruction. British Journal of Educational Psychology, 62 (2), p.233–246, https://bpspsychub.onlinelibrary.wiley.com/doi/abs/10.1111/j.2044-8279.1992.tb01017.x
Clark, R.C., Nguyen, F. and Sweller, J. (2006). Efficiency in learning: evidence-based guidelines to manage cognitive load. San Francisco: Pfeiffer, https://bpspsychub.onlinelibrary.wiley.com/doi/abs/10.1111/j.2044-8279.1992.tb01017.x
Marzano, R.J., Gaddy, B.B. and Dean, C. (2000). What works in classroom instruction. Aurora, CO: Mid-continent Research for Education and Learning, https://bpspsychub.onlinelibrary.wiley.com/doi/abs/10.1111/j.2044-8279.1992.tb01017.x
Sweller, J. (1988). Cognitive Load during Problem Solving: Effects on Learning. Cognitive Science, 12, p.257-285, https://mrbartonmaths.com/resourcesnew/8.%20Research/Explicit%20Instruction/Cognitive%20Load%20during%20problem%20solving.pdf
Wenger, S.K., Thompson, P. and Bartling, C.A. (1980). Recall facilitates subsequent recognition. Journal of Experimental Psychology: Human Learning and Memory, 6 (2), p.135-144, https://journals.sagepub.com/doi/abs/10.1111/j.1745-6916.2006.00012.x?journalCode=ppsa
https://journals.sagepub.com/doi/abs/10.1111/j.1745-6916.2006.00012.x?journalCode=ppsa
Choi, S., Yuen, H. M., Annan, R., Monroy-Valle, M., Pickup, T., Aduku, N. E. L. & Ashworth, A. (2020). Improved care and survival in severe malnutrition through eLearning. Archives of disease in childhood, 105(1), 32-39. Doi:10.1136, https://pubmed.ncbi.nlm.nih.gov/31362946/
Chyung, Y. C. (2007). Learning object-based e-learning: content design, methods, and tools. Learning Solutions e-Magazine, August 27, 2007, 1-9. https://www.elearningguild.com/pdf/2/082707des-temp.pdf.
https://dobusyright.com/druckers-six-ideas-about-knowledge-work-environments-dbr-031/
Doolani, S., Owens, L., Wessels, C. & Makedon, F. (2020). vis: An immersive virtual storytelling system for vocational training. Applied Sciences, 10(22), 8143, https://ouci.dntb.gov.ua/en/works/l1mwRvyl/
Gavarkovs, A. G., Blunt, W. & Petrella, R. J. (2019). A protocol for designing online training to support the implementation of community-based interventions. Evaluation and Program Planning, 72(2019), 77-87, https://psycnet.apa.org/record/2018-61336-010
Kraiger, K. & Ford, J. K. (2021). The science of workplace instruction: Learning and development applied to work. Annual Review of Organizational Psychology and Organizational Behavior, 8, 45-72.
Longo, L. & Rajendran, M. (2021, November). A Novel Parabolic Model of Instructional Efficiency Grounded on Ideal Mental Workload and Performance. International Symposium on Human Mental Workload: Models and Applications (pp. 11-36). Springer, Cham, https://www.researchgate.net/publication/354893819_A_novel_parabolic_model_of_instructional_efficiency_grounded_on_ideal_mental_workload_and_performance
Merrill, M. D. (2018). Using the first principles of instruction to make instruction effective, efficient, and engaging. Foundations of learning and instructional design technology. BYU Open Textbook Network. https://open.byu.edu/lidtfoundations/using_the_first_principles_of_instruction
Rea, E. A. (2021). "Changing the Face of Technology": Storytelling as Intersectional Feminist Practice in Coding Organizations. Technical Communication, 68(4), 26-39, https://g.co/kgs/XAhhL77
Schalk, L., Schumacher, R., Barth, A. & Stern, E. (2018). When problem-solving followed by instruction is superior to the traditional tell-and-practice sequence. Journal of educational psychology, 110(4), 596, https://psycnet.apa.org/record/2017-57179-001
Shipley, S. L., Stephen, J. S. & Tawfik, A. A. (2018). Revisiting the Historical Roots of Task Analysis in Instructional Design. TechTrends, 62(4), 319-320. https://doi.org/10.1007/s11528-018-0303-8
Smith, P. L. & Ragan, T. J. (2004). Instructional design 3rd Ed. John Wiley & Sons, https://www.amazon.com/Instructional-Design-Patricia-L-Smith/dp/0471393533
Willert, N. (2021). A Systematic Literature Review of Gameful Feedback in Computer Science Education. International Journal of Information and Education Technology, 11(10), 464- 470, https://www.researchgate.net/publication/354317969_A_Systematic_Literature_Review_of_Gameful_Feedback_in_Computer_Science_Education