When performance isn’t learning: What ESL teaching, academic support, and AI have taught me about how students actually learn
“What does it really mean to learn something?”
Early in my education career, I would have answered that question with far more confidence than I have now. Learning felt measurable and objective: if students improved their scores or reached their target band score, the instruction had worked and they had learned something. Over time, and across very different teaching and academic support roles, I’ve come to realize how misleading that definition can be. Students can pass, progress, and move on while still struggling to apply what they supposedly learned in real situations.
This blog reflects on how my understanding of learning has evolved from my early work as an ESL teacher to my current role in academic support. Drawing on evidence-based learning theories, I explore why performance is such a tempting proxy for learning, why certain learning myths persist, and why tools like AI can be harmful when introduced at the wrong stage. Ultimately I believe that instructional design decisions must be grounded in how learning actually works, instead of following trends, intuition, or chasing short-term outcomes.
Learning as test performance: Lessons from ESL teaching
My career in education began as an English as a Second Language (ESL) teacher working primarily with high school students and recent graduates preparing for the IELTS exam. For most of my students, IELTS was a gatekeeper. A specific band score meant access to overseas universities, scholarships, and future opportunities. In that context, learning was framed narrowly and pragmatically: memorizing vocabulary, internalizing grammar rules, learning test structures, and practicing strategies designed to maximize exam performance.
By those measures, many students were successful. They achieved their target scores and moved on. However, once they entered English-speaking academic or social environments, many struggled to communicate confidently, participate in discussions, or write effectively beyond formulaic responses. Despite years of studying English, their ability to use the language flexibly was limited. They had learned how to perform English for an exam, but not how to use English as a communicative tool.
This experience aligns with research showing that performance on assessments does not necessarily indicate durable learning or transfer (Lovett et al., 2008). Across learning theories, learning is consistently defined as an enduring change in knowledge, skills, or beliefs resulting from experience, not short-term exposure or repetition. Learning is not always immediately visible and cannot be inferred from performance alone. Without opportunities to apply knowledge meaningfully, engage socially, and receive feedback in authentic contexts, “learning” often fails to transfer.
What learning theories help explain these experiences?
Social cognitive theory offers a useful lens for understanding this pattern. Many of my ESL students were driven almost entirely by extrinsic motivation: scores, acceptance letters, and institutional requirements. Once those incentives disappeared, engagement often dropped sharply. Bandura’s concept of self-efficacy is particularly relevant here. Students who technically succeeded on exams frequently doubted their ability to communicate in real contexts, which reduced risk-taking and persistence (Bandura, 1986). I see this same pattern in academic support settings today, where students often ask, “What do I need to do to achieve this score?” rather than focusing on understanding or skill development. Modeling, guided practice, and experiences of meaningful success are therefore critical instructional strategies grounded in social cognitive theory.
Sociocultural theory further clarifies why language learning cannot be reduced to memorization or individual cognition. Learning is shaped by social interaction, cultural tools, and participation in meaningful practices. Without opportunities to use language in authentic academic and social contexts, my ESL students’ knowledge stagnated and they were unprepared for authentic English interactions. In academic support, I frequently work with students who understand content conceptually but struggle to participate in the discourse of a discipline, whether that involves academic writing, collaborative work, or critical discussion. Learning becomes durable only when students are supported in engaging with the norms and practices of their learning environment.
Cognitive load theory has become especially relevant in my current role as an academic coordinator, advisor, and lecturer. In our department’s foundational Learning to Learn (L2L) course, AI tools have been introduced to support basic academic skills such as sentence and paragraph writing. While well intentioned, this design often overwhelms students who are still developing foundational literacy skills. These students are expected to interpret assessment tasks, generate language, prompt AI tools, evaluate output quality, identify inaccuracies, and make ethical decisions about use simultaneously. For novice learners, this creates an excessive cognitive load that overwhelms working memory and interferes with schema construction (Kirschner & van Merriënboer, 2013). Without sufficient prior knowledge, students are not equipped to evaluate AI-generated content meaningfully, and the tool risks replacing learning rather than supporting it.
Academic support and the need for coherent learning design
My work in academic support has also highlighted how the absence of a coherent development and evaluation process can undermine learning. The L2L course currently lacks a clear learning flow and has not been developed through a structured cycle such as ADDIE or systematically evaluated using frameworks like Kirkpatrick’s Four Levels of Evaluation. Without needs analysis or iterative evaluation, instructional decisions have become reactive rather than intentional, and the course has accumulated piecemeal contributions over several years.
The rapid inclusion of AI tools in the course appears to be driven more by perceived innovation than by alignment with learner readiness or subsequent course demands. Research on instructional design consistently warns against introducing complex tools before foundational knowledge is established, particularly for beginner learners (Kirschner & van Merriënboer, 2013; Lovett et al., 2008). From a learning science perspective, this reinforces the importance of sequencing instruction appropriately, managing cognitive load, and designing with learner development in mind.
Myths about learning
One of the most persistent myths I encounter in both teaching and academic support is the belief that instruction should be tailored to individual learning styles. Early in my ESL career, we introduced learning styles to students and attempted to design activities around them. I remember spending a surprising amount of time developing “kinesthetic” vocabulary activities and encouraging students to move around the classroom to memorize words. Although intuitively appealing, there is no credible empirical evidence that matching instruction to learning style preferences improves learning outcomes (Kirschner & van Merriënboer, 2013; Veritasium, n.d.). This myth is harmful because it encourages fixed beliefs about learning and distracts from strategies that actually support learning, such as practice, feedback, and meaningful engagement. I still think it’s pretty funny that at one point we were telling students to close their eyes while they practiced conversations with each other though.
Another damaging misconception is the assumption that poor performance reflects low ability. Research shows that performance is shaped by instructional design, feedback, motivation, and cognitive load rather than by ability alone (Lovett et al., 2008). Across both ESL teaching and academic support, I have seen how system-level design decisions often influence outcomes more than individual learner capacity. In courses like L2L, confusion is often attributed to students rather than to fragmented design or unrealistic expectations. While there is rarely a single cause, these experiences have strengthened my belief that improving learning outcomes requires addressing instructional structures rather than defaulting to deficit explanations.
Conclusion: Learning beyond performance
Across my experiences in ESL teaching and academic support, I have learned that performance is an unreliable indicator of learning. Test scores, polished submissions, and technically correct outputs often conceal fragile understanding and skills that fail to transfer beyond specific tasks or contexts. When instruction prioritizes visible productivity or short-term outcomes, it becomes easy to mistake activity for learning.
These experiences have reinforced the importance of grounding instructional design decisions in evidence rather than trends or intuition. Poor learning outcomes are rarely the result of individual student deficits alone; they are more often shaped by instructional sequencing, cognitive load, learner readiness, and opportunities for meaningful participation. Designing for learning means accepting that real learning is slower, less visible, and more complex than performance metrics suggest. It is also the only way to support learning that lasts beyond the assessment.
Thanks for reading, and I’m looking forward to learning more about learning with everyone!
References
References
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Prentice Hall.
Kirschner, P. A., & van Merriënboer, J. J. G. (2013). Do learners really know best? Urban legends in education. Educational Psychologist, 48(3), 169–183. https://doi.org/10.1080/00461520.2013.804395
Lovett, M. C., Meyer, O., & Thille, C. (2008). The Open Learning Initiative: Measuring the effectiveness of the OLI statistics course in accelerating student learning. Journal of Interactive Media in Education, 2008(1). https://doi.org/10.5334/2008-14
Veritasium. (n.d.). The myth of learning styles [Video]. YouTube. https://www.youtube.com/watch?v=rhgwIhB58PA