Page 164 - Izobraževanje v dobi generativne umetne inteligence

Saša Mladenović

the importance of guided interaction between the learner and a more knowl-
edgeable other—traditionally a teacher or peer. While Generative AI can act
as a tool to extend the ZPD by providing timely assistance, it cannot fully
replace the dynamic and responsive interaction that characterises effective
scaffolding. Teachers and peers bring context, empathy, and the ability to
challenge misconceptions, elements that Generative AI still cannot replicate.
Furthermore, the inaccuracies in Generative AI’s responses, stemming from
its misunderstanding or limitations, pose a significant risk of misguiding stu-
dents if these errors go unchecked.
Jerome Bruner’s theory of discovery learning (Ozdem-Yilmaz & Bilican,
2020) advocates for teaching methods that encourage students to construct
new ideas based on their existing knowledge. Generative AI has the potential
to facilitate discovery learning by generating examples, explanations, or al-
ternative perspectives. However, the risk lies in students passively accepting
the AI’s outputs as authoritative answers, bypassing the critical thinking and
inquiry essential for deep learning. Unlike a teacher or peer, Generative AI
cannot discern whether a student has fully grasped a concept or is simply
replicating its outputs without understanding. We are experiencing the Chi-
nese room argument (Cole, 2024).
Seymour Papert’s constructionism (Papert, 1984, 2020; Stager, 2005) builds
on Piaget’s ideas, emphasising learning through making and hands-on activ-
ities. Generative AI can support constructionist approaches by helping stu-
dents brainstorm, simulate, or design creative projects. However, its lack of
awareness about individual learners’ contexts or objectives means it cannot
offer the kind of meaningful feedback or encouragement that a teacher or
peer might provide during project-based learning.
A critical challenge with GEN-AI lies in its tendency to occasionally pro-
duce incorrect or misleading information due to its reliance on probabilistic
models rather than genuine understanding (Garry et al., 2024). Teachers and
peers, while not infallible, are better equipped to acknowledge the limits of
their knowledge and collaboratively seek accurate answers. Generative AI, on
the other hand, can present inaccuracies with unwarranted confidence, po-
tentially leading students astray if they lack the skills to critically evaluate its
outputs.
In summary, while Generative AI has the potential to enrich the learning
experience, it cannot substitute the nuanced, responsive, and context-aware
roles of teachers and peers. Its integration into education must be carefully
guided by pedagogical principles that emphasise critical thinking, discovery,
and active participation, ensuring that students engage with Generative AI

164