Dr. Barry Leckenby
GenAI is the latest in a long line of technologies to prompt debate about their effects on thinking, learning, and professional practice.
For educators and learning designers, the immediate challenge is practical: GenAI changes how learners initiate work, draft responses, and access feedback, which in turn reshapes what instructional support is needed and where assessment must focus. Compared with earlier tools such as calculators and search engines, GenAI extends beyond information retrieval by enabling conversational interaction, rapid drafting, and context-sensitive feedback—capabilities that can support performance while also introducing new risks.
As with prior technologies, benefits and costs coexist. GenAI tools can support writing productivity by assisting with idea development, content generation, and revision. By summarizing and organizing text, GenAI may reduce initial information-processing demands and support early drafting—particularly for learners who struggle with task initiation or low confidence. GenAI may also help reduce teacher workload through scalable feedback and automated support mechanisms, provided these are integrated responsibly.
Cognitive Effort Across Workflows
Studies since the release of widely used GenAI tools in late 2022 have produced a growing body of empirical evidence relevant to cognitive development and task performance in learning and knowledge-work contexts. In education, GenAI adoption has prompted renewed attention to how cognitive effort is distributed across a workflow, shifting emphasis from initial drafting and information organization toward verification, evaluation, and integration of generated outputs.
While cognitive development and productivity gains are two of the most anticipated benefits of GenAI, there have also been concerns about its possible drawbacks, including over-reliance on it as an epistemic authority and use of GenAI causing a decrease in comprehension, practical application, analysis, evaluation and creative synthesis.
More recent empirical work, however, suggests a more nuanced pattern has emerged in which effort is often redistributed from initial production toward later-stage monitoring, evaluation and revision—particularly where users actively verify and improve outputs.
Across several studies, GenAI support has reduced effort devoted to early-stage work such as initial drafting and information organization, while increasing the importance of later-stage evaluation, verification, and synthesis to ensure quality and avoid error. The empirical data also indicates GenAI assists the lower end of competency (novices) more than the higher end of competency (experts). Performance gains also appear heterogeneous: Lower-performing participants often show larger improvements in task time and output quality measures, whereas higher-performing participants may primarily gain efficiency.
Strategic Prompting to Improve Performance
Prompt engineering is an applied skill set combining domain knowledge, critical evaluation and iterative refinement of prompts through testing and revision. Empirical HCI research shows non-experts often struggle to translate goals into effective prompts and strategies such as adding examples, adopting more structured prompt formats and iteratively rephrasing requests can improve prompt effectiveness.
Prompt pattern catalogs have proposed reusable templates intended to scaffold prompt construction. Chain-of-thought is a prompting strategy for more complex problem-solving, in which the steps of reasoning are included in the prompt and can improve performance on multi-step reasoning tasks in some settings. In-context learning (e.g., few-shot demonstrations) is a common technique that can improve performance on new tasks without parameter updates, although effectiveness varies by task and prompt design.
While in-context learning prompts are an important emergent property of GenAI, improving performance by carefully designed input prompts indicates learning designers and educators could use prompting strategies as metacognitive scaffolds.
A promising approach is to develop and refine steering prompts collaboratively with teachers and subject-matter experts to align prompts with disciplinary standards and learner needs. The roles of teachers, subject matter experts and GenAI will become increasingly complementary and collaborative. Their domain expertise is vital for prompt optimization, as it will provide the most effective steering prompts.
Harnessing the potential of GenAI in productive and ethical ways requires a systematic approach to prompt engineering. Common prompting approaches, such as zero-shot prompts, few-shot demonstrations and chain-of-thought prompting, can improve performance for some tasks but vary in reliability across contexts. Chain-of-thought prompting can sometimes improve multi-step reasoning without model retraining, potentially reducing reliance on task-specific fine-tuning.
Task Delegation
Apart from prompt optimization, the other most important new cognitive tasks occur at the beginning and end of generating automated output. Task delegation needs to occur before prompt optimization, which means understanding what tasks should be delegated to GenAI and what tasks should not be automated.
After automated output has been generated, it will need to be evaluated. Evaluating automated output is not necessarily straightforward and requires the application of the critical thinking skills of evaluation, comparison, and analysis, as well as navigating complexity with domain expertise, which novice learners will not necessarily have at hand.
The successful leveraging of GenAI tools and skills in knowledge task flows depends on completing task delegation, prompt optimization, and generated output evaluation.
Task delegation can be developed as a learner capability, but typically benefits from explicit instruction and scaffolding to help novices decide what to automate and what requires human judgment. To foster a collaborative approach, prompts should be collected into libraries of prompt templates and discussed at team meetings. For interdisciplinary teams, prompt optimization will foster interpersonal communication and provide a more holistic approach to problem-solving. A major value proposition of GenAI may lie in human–AI collaboration that supports ideation and creative recombination, particularly when users actively steer and curate outputs.
Cognitive Rebalancing
Cognitive rebalancing can be characterized in three ways. Effort shifts:
- From information gathering to information verification at the level of knowledge and comprehension
- From problem solving to response integration at the level of application
- From task execution to task stewardship at the levels of analysis, synthesis, and evaluation
While automation can reduce production effort, increased demands on monitoring and evaluation may impair situational awareness, particularly in complex or opaque systems. Confidence further moderates these dynamics. Greater task confidence supports effective delegation and stewardship, whereas lower self‑confidence may increase reliance on GenAI and reduce critical engagement. Importantly, qualitative evidence indicates that users enact critical thinking when reflecting on and improving automated outputs, suggesting that individuals who already engage in reflective practice tend to maintain such engagement when using GenAI tools.
Because cognitive resources saved through automation are not reliably redirected to more demanding tasks, instruction and feedback must explicitly accompany GenAI use to support critical thinking development.
Decades of educational research demonstrate that targeted instruction combined with timely feedback is among the most effective ways to enhance learning. Human‑factors research further shows that automation can make simple tasks easier while exacerbating difficulty in complex tasks, underscoring the need for deliberate task allocation between humans and GenAI.
Scaffolding therefore plays a central role in supporting cognitive development. Learners with differing levels of expertise and confidence require tailored prompts, examples, and reflective guidance to avoid over-reliance and to sustain epistemic agency.
Task characteristics also matter in professional education contexts: GenAI assistance has been shown to improve performance on structured assessments such as multiple‑choice questions, while offering limited or negative benefits for open‑ended tasks such as essays, particularly for high‑performing students. GenAI can provide valuable initial scaffolding through templates and examples, but learners must still engage in problem framing, comparison of alternatives, and evaluation.
Implementation Framework
To operationalize the effective and ethical integration of GenAI into knowledge‑task workflows, a structured implementation framework is recommended. The framework is organized around four components: inputs, process, outputs, and assessment, enabling systematic adoption across educational and workplace settings.
Inputs
Effective implementation requires the provision of shared cognitive and evaluative resources. These include:
- Prompt libraries, comprising curated prompt templates aligned to task types, domains, and levels of learner expertise, to support prompt optimization and reduce trial‑and‑error overhead
- Evaluation checklists, designed to guide verification of factual accuracy, relevance, and coherence, and evaluation of bias and contextual appropriateness of GenAI outputs
- Risk rubrics, which articulate acceptable use boundaries based on task stakes, domain sensitivity, and potential consequences of error, thereby supporting informed task delegation
Process
The framework follows an iterative, guided workflow that embeds critical thinking before, during and after GenAI interaction:
- Task delegation workshops support learners and knowledge workers in determining which components of a task should be automated and which require human judgement and accountability
- Guided prompting cycles involve iterative prompt design, testing, critique, and refinement, ideally supported by teachers, experts or peers
- Evaluation and reflection activities require users to interrogate automated outputs against established criteria, encouraging comparative analysis, judgement and metacognitive awareness
Outputs
Implementation produces tangible artifacts that make cognitive processes visible and assessable, including:
- Prompt logs documenting iterations, design decisions, and rationale
- Evaluation notes capturing verification steps, identified limitations, and corrective actions
- Revised drafts or final outputs that demonstrate informed integration of GenAI contributions rather than uncritical adoption
Assessment
Assessment should focus on the quality of cognitive engagement rather than the mere presence of GenAI use. Rubrics should evaluate:
- Verification: The extent to which automated outputs are cross‑checked, contextualized, and validated
- Judgment: The quality of evaluative reasoning, comparison of alternatives, and task stewardship
- Transparency of reasoning: Clarity in documenting how GenAI outputs were prompted, assessed, modified, or rejected
By foregrounding task delegation, prompt optimization, and output evaluation as learnable and assessable practices, this framework supports the rebalancing of cognitive load toward higher‑order thinking while mitigating over-reliance, premature convergence, and performance degradation. It provides a practical structure for leveraging GenAI as a collaborative tool that enhances both cognitive development and productivity.
Explore AI & Learning Design
Don’t miss the Learning Guild’s AI & Learning Design online conference, August 26-27, 2026. Kicking off the two-day event, keynoter Beth Ridley describes what makes learning professionals indispensable. Next, we’ll explore how AI is changing workflows, demanding complex choices about what to automate and where human expertise is still needed. The six sessions move beyond exploring AI-driven tools into probing decisions about speed vs. quality, trust, accessibility, and ensuring that learning experiences are engaging and meaningful.
Over two days, you’ll explore ways to use AI safely in regulated or secure contexts, streamline development, address performance gaps, retain the human connection, and more. Each day, you can join our new ThinkSpaces to chat with speakers and other attendees at virtual roundtable discussions.
Registration is open—and the online conference is free for Professional members!
Image credit: iMrSquid
