Framing Generative AI applications as tools for
cognition in education
Enmarcando las aplicaciones de IA
generativa como herramientas para la cognición en educación
Dr. Marc
Fuertes-Alpiste. Profesor agregado. Universitat de
Barcelona. España
Recibido:
2024/04/06 Revisado 2024/04/29 Aceptado: :2024/07/01 Online First: 2024/07/05 Publicado: 2024/09/01
Cómo citar este artículo:
Fuertes
Alpiste, M. (2024). Enmarcando las aplicaciones de IA generativa como
herramientas para la cognición en educación [Framing
Generative AI applications as tools
for cognition in education]. Pixel-Bit. Revista De Medios Y Educación, 71,
42–57. https://doi.org/10.12795/pixelbit.107697
ABSTRACT
Generative AI applications
enable different useful functions for learning based on the generation of
content. This paper aims to offer a theoretical framework to understand them as
tools for cognition (TFC), framed in the perspective of sociocultural theory
and activity theory and distributed cognition. This perspective exemplifies how
thought is not only packaged inside the individual's mind, but is distributed
among subjects, objects, and artifacts, where tools mediate human activity and
help in the executive functions of thought. The perspective of TFC embodies an
educational socio-constructivist vision where learners build their knowledge
with these tools, taking advantage of their affordances. It is the concept of
learning "with" technology instead of the traditional vision of
learning "from" technology, where technological applications are
limited to providing information and evaluating the students' responses.
Finally, we describe Generative AI applications as HPC following David
Jonassen's pragmatic and pedagogical criteria, i.e. the capacity of knowledge
representation of different subjects, the facilitation of critical and
meaningful thinking (based on questioning and prompting) and how they enable
complex thinking among students when used in learning tasks, only when
executive functions are on the side of the learner.
RESUMEN
Las aplicaciones de IA
generativa permiten funciones útiles para el aprendizaje basadas en la
generación de contenido. Este artículo ofrece un marco teórico para entenderlas
como herramientas para la cognición (HPC), basado en la perspectiva de la
teoría sociocultural, la teoría de la actividad y la cognición distribuida.
Esta perspectiva ejemplifica cómo el pensamiento no sólo está empaquetado
dentro de la mente, sino que se distribuye entre sujetos, objetos y artefactos,
donde las herramientas median la actividad humana y ayudan en las funciones
ejecutivas del pensamiento. Encarna una visión en la que los alumnos construyen
su conocimiento con ellas, aprovechando sus posibilidades de acción. Es la
concepción de aprender "con" la tecnología en lugar de la visión
tradicional de aprender "de" la tecnología, donde las aplicaciones
tecnológicas se limitan a proporcionar información y a evaluar las respuestas
de los estudiantes. Finalmente, describimos las aplicaciones de IA generativa
como HPC siguiendo los criterios pragmáticos y pedagógicos de David Jonassen, como la capacidad de representación del
conocimiento, la facilitación del pensamiento crítico y significativo (basado
en preguntas y prompts) y cómo permiten el
pensamiento complejo entre estudiantes cuando se utilizan en tareas de
aprendizaje, solamente cuando las funciones ejecutivas las realizan ellos.
PALABRAS CLAVES· KEYWORDS
Artificial intelligence;
cognition; learning processes; critical thinking; computer uses in education
Inteligencia artificial;
cognición; procesos de aprendizaje; pensamiento crítico; usos de los
ordenadores en educación
1. Introductión
Generative AI (hereinafter, GAI) applications have an
increasing interest in educational research and are presented as an opportunity
for personalizing learning, as a means of personal assistance, and as cognitive
supports for higher order thinking, but also as a source of ethical problems
and biases, lack of academic integrity, privacy issues and dissemination of
false information (Crompton & Burke, 2024; Mishra et al., 2024; Walter,
2024).
According to UNESCO, GAI is " (...) an artificial
intelligence (AI) technology that automatically generates content in response
to prompts written in natural-language conversational interfaces" (Miao
& Holmes, 2024, p.8). Moreover, it uses various AI technologies to create
content across various media formats (Schellaert et
al., 2023). This capacity to generate content that is plausible to people is
what makes GAI different from past AI technologies, along with the social
dimension that derives from its own interface: that is based on natural
language, and we communicate with it through chatboxes
or directly with our own voices. It appears as a human agent with whom users
relate using such a human feature as language (Author, 2018; Mishra et al.
2023).
There are several GAI applications that enable
different useful functions in education based on the generation of content
(such as ChatGPT, Copilot etc.), either to create, structure, synthesize,
reformulate texts or ideas, by students and/or by teachers, individually and/or
collectively. There are GAI applications that guide us in the search of
literature (Perplexity), in summarizing content (Scribe, Claude) in breaking
down complex tasks in steps (Goblin Tools), that help with code writing (GitHub
Copilot, AlphaCode), generate content in multiple
media (DALL-E, Sora, Synthesia), among others.
These GAI application affordances have been appraised
in the educational field. For instance, a UNESCO report from April 2023
identified its use in Higher Education as an aid in refining ideas, as an
expert tutor, etc., for students’ learning and teaching (Sabzilieva
& Valentini, 2023). Crompton and Burke's (2024) systematic research review
identified teachers' uses such as for teaching support, task automation, and for
student learning such as accessibility, explaining difficult concepts, acting
as a conversational partner, providing personalized feedback, providing writing
support, self-assessment, and facilitating student engagement and
self-determination.
Given the importance they have in the current
educational debate, this paper aims to offer a theoretical framework to
understand the GAI applications as cognitive
tools, mindtools or tools for cognition (TFC from now on)
-which stems from the vision of Gavriel Salomon, Roy D. Pea, Howard Rheingold,
David Jonassen, among other introducers of the concept at the end of the 20th
century-, framed in a perspective of sociocultural theory, activity theory and
distributed cognition. Eventually, theoretical frameworks for the integration
of GAI in education are needed to scaffold research efforts that can help
evolve these theories of AI in Education (Dawson et al., 2023), but they should
also structure a view of education that places the student at the center of the learning process, that enhances his/her
agency, autonomy in the executive functions of cognition and critical thinking,
and not to become just an artificial teacher or a tutorial-based educational
system. This, we find, would be a limiting perspective. We intend to provide
theoretical knowledge that informs how to effectively integrate GAI
applications in our educational practice. The awareness of educational
philosophies that help shape our own views provides us with a consciousness of
the best technological choices for the greatest outcomes for our learners. This
is what underlies our views of education and technology and what brings
coherence and consistency to them (Kanuka, 2008).
2. Learning with technology: the
perspective of TFC in education
The use of digital technologies in education for
enhancing teaching and learning processes has been studied since their
appearance and dissemination. Derry and Lajoie (1993), Salomon et al. (1991)
and Jonassen (1996) distinguished two ways of technology integration in education: learning from
technology and learning with technology, and as simple it may seem, the
difference is huge between these two prepositions. With the first computers, Computer-assisted
instruction (CAI) and Intelligent tutoring systems (ITS) were developed. They
followed behaviorist approaches (based on
stimulus-response and reinforcement of behavior), and
cognitivist approaches (considering how our thinking process is related to
working memory, long-term memory, schemata in our previous knowledge, memory
retrieval, and elaborated feedback and personalization of learning). This
learning approach is based on the learning from technology approach,
where technology has the role of a teacher who gives the input and the feedback,
accordingly, reproducing the traditional education approach -teacher-centered-, where the students' learning agency is low.
On the other hand, the learning with technology
approach is learner-centered, where the student uses
technology as a tool, aimed at doing something with it, not as a tutor. Thus, from
this perspective, cognition is on the side of the learner, who takes advantage
of the digital tools' affordances to do something with the tool that would not
be possible (or at least it would be harder) without it. This perspective is
rooted in the constructivist approach of teaching and learning where teachers
design student-centered activities with preestablished
pedagogical aims in which students have to construct their own learning.
According to Iiyoshi et al. (2005), this is carried
out through five different cognitive processes: information search, information
presentation, knowledge organization, knowledge integration and knowledge
generation. Derived from this, teachers can also adopt a constructionist
perspective, where kids' activity is oriented to the creation of artifacts,
unfolding their creativity (Papert, 1982), or
socio-constructivist approaches of learning, where learners interact actively
with the environment and with their peers and their educators to construct
their learning.
3. Sociocultural theory as an overarching theory
for TFC
The learning with technology perspective is
also rooted in sociocultural theory, created by Vygotsky, Leontiev and Luria during the 1920s in the Soviet Union. It
assumes that the historical development of human culture is different from
human biological evolution because it has its own rules (Vygotsky, 1978).
Cultural development is based on the use of tools, created by humans, to act on
the environment. They can be physical tools (e. g., a hammer or a screwdriver),
but also symbolic tools (e.g., language). Symbolic tools are decontextualized
from the environment – nature, biology – since the symbols they manipulate are
increasingly less dependent on the space-time context in which they are used. A
person without tools, only considering his/her biological evolution, cannot
evolve; without tools, the qualitative cognitive leaps that Vygotsky relates to
the transition from elementary cognition – that which is related to the primary
and natural – to higher cognition – the "cultural" and the social –
do not occur (Wertsch, 1995).
This influence also has an effect on building new
tools that will affect the physical and cultural environment that will, in
turn, also affect culture. This means that these tools make us, and they are
not just a matter of the present time, but they come from our ancestors, and
they will have an impact on the future. It focuses on activity as a culturally
mediated action (Engeström & Sannino, 2021). From
this perspective, the question revolving around whether to use or not to use
digital tools in our human activity -and here we really mean "in
education"-, is a false debate as tools are part of us. Without human
beings there is no human culture, but without culture -and cultural tools-
there are no human beings as we know them.
Sociocultural theory is based on culturally mediated
action and is normally represented in the form of a triangle with three actors
in each vortex (see Figure 1): In the lower part, the Subject and the Object,
there is no mediation. The Medium in the upper vortex is the artifact
(the tools) that mediates the subject's action on the object (the environment,
the other subjects). This is a representation of what is often called the first
generation of Activity Theory (Engeström &
Sannino, 2021).
Figure 1
The triangle of culturally mediated action
In a second generation of Activity theory, the
interactive triangle in Figure 1 was expanded to represent Leontiev's
(1978) idea of activity (not just a sole action), which helps represent a
larger mediation (Cole & Engeström, 1993). In Figure
1, the interaction is based on the individual action of the subject, mediated
by the Medium, but in the second generation (see Figure 2), the main unit of
analysis is the activity, that involves a community (the social dimension),
carried out following a set of rules (on the left side). In the right vortex of
the base, we find the "division of labor"
where the activity is divided between participants and tools (Bakhurst, 2009).
Figure 2
A representation of the second generation of Activity
Theory
What we want to stress is that this mediation places
people's cognition within social and cultural contexts of interaction and
activity (Salomon, 1993). We can say that this triangle is a map of a
distributed cognition, between people, community, and cultural tools and
artifacts. In this action system (Figure 1) and activity system (Figure 2), our
mind is distributed with the tools -creating a distributed cognition- that we
use to act on the environment (Cole & Engeström,
1993). This is why it is important to understand how these tools can offer
different forms of mediation, and GAI applications can have a role in human
action and collective activity as other digital tools also have.
4. Distributed cognitions with tools
With the first computer applications it was possible
to imagine a collaboration between users and digital tools for a better
performance (Salomon, 1993). These tools help us in tasks, guiding the activity
and augmenting our capabilities, but also affecting our cognition, when we
internalize these new forms of action (Vygotsky, 1978). The concept of
distributed cognition was studied from different perspectives (see Salomon,
1993), and it implies that the activity needs to adapt the means to the ends. It
leverages the tools' affordances, which are the actual and perceived properties
of tools (Gibson, 1979), but more particularly as possible actions that users
can develop with tools. It is not just what the user perceives in a tool that
will allow certain operations in a social activity or in an individual action,
but all the possible actions we could undertake with a tool (Norman, 1999) (for
instance, with GAI applications). Affordances are not always evident and therefore
must be learned (Gibson, 1989).
It is worth mentioning that these tools contain an
intelligence within their designs, what Lave (1988) describes as
"mythical" artifacts because they are already a part of our
consciousness and we are not aware of what they are, they become invisible,
like a speedometer in a car or a house thermostat. According to Dubé &
McEwen (2017), when we are using tools, we perceive the symbols for the
communication of affordances (for example, identifying a chatbox,
a history log in certain GAI applications), the actual affordances themselves,
and affordances that may be perceived by users based on their cognitive
abilities.
As Clark and Chalmers (1998) described, there is an
active effective externalism of our cognition where the environment has an
active role in our internal cognitive processes. When we use digital tools to
manipulate information through their options, they afford cognitive processes,
augmenting our cognitive capacities, such as when a player rotates Tetris
figures to make them fit in the spaces using the videogame affordances and not
just by rotating them mentally. The rotation affordance lets the player visualize
more quickly the action to be undertaken than if it had to be done just
mentally. As described by Kirsh and Maglio (1994), these are the
"epistemic actions", that help uncover information that is hard to
process and which contribute to simplify problem-solving tasks. They are
different from pragmatic actions, which help reach a physical goal.
Technologies are part of activity systems (Scribner
& Cole, 1981), so when taking GAI applications into account, we must elicit
the activities they afford. The nature of the activity will affect the
cognitive processes, not just the tool per se (Salomon & Perkins, 2005).
Sharples (2023) identified possible interactive and social learning uses
between students and GAI applications, e.g., to explore possible scenarios, as
a Socratic "opponent", as a co-designer, as a help for data or
information interpretation, and so on.
When performing an activity, cognition may be
distributed between the person and the tool, and here we find two quite
different perspectives (Salomon et al., 1991). Firstly, the systemic one, which
is an aggregated performance of the person-tool (this is Pea's position). Distributed
cognition is not a matter of sharing or reallocating intelligence between mind,
context, and tools, but of stretching intelligence during the activity (Pea,
1993). Secondly, the analytic perspective considers the specific contributions
of the person and the tools (that would be Salomon's choice), where the person
has a predominant role, most importantly because the tool really does not
understand anything during the distributed cognition, just the individual does.
When talking about GAI applications such as ChatGPT, this has been a critic
from the beginning as Chomsky et al. (2023) have pointed out. The system does
not understand language, so it is not real intelligence that is stretching over
between person and tool. But it does not mean that these tools are not helping
in off-loading a cognitive load in a task and that the tools themselves have
intelligence of a social origin, as cultural tools.
During a cognitive partnership, the person will have the
effects of using tools in the cognition, at least in three forms: the effects
with technology, where using a technology enhances the cognitive processing
and performance (this is an augmentation), the effects of technology
where technology use leaves a "cognitive residue in the form of improved
competencies, which affect subsequent distributed activities" (Salomon,
1993, p.123). And finally, we find the effects through
technology, where technology use does not just augment our intellectual
processing capacity, but it also reorganizes it (Salomon & Perkins, 2005).
David Perkins advocated for the distributed cognition
perspective with the concept of the "person-plus" as opposed to the
"person-solo" when dealing with tasks and activity. The
"person-plus" takes advantage of the tools that help with the
cognition, just as a student's learning is not just in what is inside his /her
head but also in the "student-notebook" system (Perkins, 1993).
When dealing with collaboration between humans and
artificial agents, the concept of integrated hybrid intelligence
emerges. It refers to increased effectiveness in human activity. However, we
find that it only appraises the side of the augmentation of human capabilities
(the quantitative aspect, and not the qualitative one) (Akata
et al., 2020; Järvelä et al., 2023). For instance, according to Holstein et al.
(2020), this collaboration in performance and mutual learning happens for an
augmentation in goals, perception, action, and decision-making. We think it
does not consider how this intellectual partnership happens within an action or
an activity system where interaction is not limited to augmentation but as a
human activity reorganization.
5. Framing GAI apps as TFC
Understanding digital tools as TFC means adopting a
perspective of learning with tools, and not just from them. They
have been called "tools for thought" (Rheingold, 1985),
"cognitive tools" (Pea, 1985; Salomon, 1993), "tools for
cognition" (TFC) (Pea, 1993), "Mindtools" (Jonassen, 1996), and
"psychological instruments" (Kozulin,
2000). In the case of this paper, we chose the term "TFC" as these
tools are not intelligent and because they are tools for a purpose of an
activity, which engage our cognition. "Cognitive tools" would mean
that they have been already created for cognition, but many of these applications
have been created for open purposes, which is the case of GAI apps. They can
become TFC depending on the educational purpose we give them (the learning with
perspective). However, the meaning of the term is abstract and with a diversity
of views (Kim & Reeves, 2007).
There are several classifications of digital tools as
TFC (or mindtools) regarding their nature and the nature of actions they afford
(Jonassen et al., 1998) or the cognitive processes they allow (Iiyoshi et al., 2005). We find Kim & Reeves' (2007)
classification interesting, which considers (1) the kind of knowledge they
process -albeit general, domain-generic or domain-specific-, (2) the level of
interactivity between user and tool, and (3) the kind of representation they
allow -from concrete (isomorphic) to abstract (symbolic)-.
All these classifications are quite dated, and they do
not include GAI apps. The closest thing to AI applications we find in these
taxonomies would be expert systems, but they are different to GAI applications
as they function with student models and scaffold students' learning depending on how they are progressing according to such models. By
contrast, GAI apps are based on content generation and prompting. It may be
similar when users query a data base, but in this case, the database limits
itself to a response that is already found in its memory, whereas the GAI app
queries receive new responses generated automatically live, affording a chat.
In any case, GAI applications can perform executive functions as creators of
content, and students can just sit back and let the application do all the
writing for them. But, as TFC, they must be used in a way where executive
functions are on the side of the learner's cognition. And they have the unique
feature that can be used in general, generic and specific knowledge domains,
and their representation capacity can be either concrete or abstract.
According to Jonassen (1996) there are three basic
practical criteria that TFC (mindtools, in his terms) should have, plus
six pedagogical criteria. We discuss these criteria to frame GAI apps as TFC
below:
5.1. Practical criteria
1. That they are computer-based:
This first feature seems out of date as almost three decades
have passed and digital tools are now pervasive in our daily life activities.
Indeed, GAI apps are digital tools that are based on Large Language Models. We
can access them through cloud-based applications, using computers, smartphones
and tablets.
2. That they are available as (digital) tools:
Normally, GAI applications are available after having
registered to a platform. These tools have taken the form of chatbots with a chatbox interface and allow natural language use. This kind
of interface is usable, ergonomic and very familiar for students as many of the
communication apps are based on such interfaces.
3. That they are affordable for the public domain:
GAI tools are spreading in large numbers (Alier et
al., 2024). OpenAI claims to have the mission to “ensure that artificial
general intelligence benefits all of humanity” (OpenAI, 2024), and despite not
being entirely sure how sincere this intention is, for now they offer some of their
features for free, although others are licensed. This also happens with other
GPT tools, many of them free for teachers and educators.
5.2. Pedagogical criteria
4. Knowledge representation capacity:
GAI apps have the capacity of writing with a plausible
appeal as their special feature. They can summarize, translate, paraphrase in
different styles and speech registers, and generate all kinds of content from
their database, or introduced by the user. People can use them for tasks
related to information, which are very common in educational settings. When
dealing with information, there is knowledge, the representation of it,
retrieval of information and construction (Perkins, 1993).
5. Generalizable to several knowledge domains:
According to Schellaert et
al. (2023), there are three unique properties of GAI applications. First, there
is flexibility in the input-output diversity and multimodal capacity of these
systems. Second, generality, as they can be applied to a wide range of tasks.
And finally, originality, as they afford the generation of new and original
content. They have a knowledge base that can be very wide like the one in
ChatGPT. But many of these apps can be customized adding a knowledge base.
6. They foster critical thinking:
One of the dangers that the educational community has
placed on AI tools is that they can be used in a non-ethical way (Crompton
& Burke, 2024; Sharples, 2023). There is a fear that the student can, and
will, cheat with them and will submit assignments that have been written
directly by GAI applications. If this happens, then, the student cognition engagement
in the task is low (in that the student reviews minimally what the GAI app has
generated) or null. This has nothing to do with critical thinking, on the
contrary, it promotes a superficial processing of information.
But GAI apps have the affordance of promoting critical
thinking when being used as TFC. It is important to point out that the aim is
to engage and enhance the learners' cognition with this partnership. Cognition
engages when learners develop an activity that entails thinking in meaningful
ways, to access, represent, organize and interpret information, helping
students think for themselves, making connections and creating new knowledge
(Kirschner & Erkens, 2006).
To this aim, the use of GAI applications as TFC must
be activity oriented, thus goal oriented. This means that the learner must
provide prompts to the tool and critically refine them to get the best results.
Eager & Brunton (2023) proposed a prompting process that begins with a goal
setting, a specification of the form the output should take, writing the actual
prompt and testing and iterating until a desired output is obtained.
Providing good prompts - prompt engineering- is a
desirable skill for AI literacy and to leverage GAI apps for learning because
it requires a logical chain of reasoning (Knoth et al., 2024). When having to
give the app an input for prompting a response, the student must use simple and
clear language, giving examples to model the desired outcome, provide context,
and most importantly, refine and iterate, when necessary, also maintaining
ethical and responsible behavior (Miao & Holmes,
2023).
It is important to foster AI literacy about prompt
engineering to understand what works best to generate adequate responses, and
this entails thinking critically and creatively. This can be done with zero-shot
prompts -when we ask the GAI application a single and generic prompt to obtain
a generic answer-, or a few-shot prompts -where the user refines the
prompts with examples that the answer should contain until receiving an
adequate answer-. These are input-output prompts, but we can also develop chain-of-thought
prompts where we ask the program to explain the output step-by-step so that it
can be thoroughly assessed and help cultivate critical thinking in education
(Walter, 2024).
Moreover, when prompting, there is a need for the user
to possess a fair amount of content knowledge in order to assess the obtained
responses or outputs, which is in addition to the critical thinking skills required
to verify them and the necessary iteration of the prompt refinement (Cain,
2024; Eager & Brunton, 2023). This is the knowledge (skills, attitudes and
dispositions) required to learn facts, and it is the knowledge base for
critical thinking and creative thinking (Jonassen 1996; Perkins, 1993). Complex
thinking skills such as problem-solving, design and decision-making, which are
executive functions of cognition, can be supported by tools (Perkins, 1993).
7. They afford a transfer of learning:
According to Jonassen (1996), the transfer of learning
is directly related to problem solving, so any thinking that GAI applications
promote facilitates problem solving and transfer of learning. TFC are
generalizable tools and can be used in different settings to facilitate
cognition (Kirschner & Erkens, 2006). GAI apps are not domain dependent, so
a transfer of skills can take place between domains, and prompting to interact
with these systems can be transferred to a variety of knowledge domains
(Walter, 2024).
When we speak of the effects with and the effects of technology (Salomon, 1993), we mean that it is desirable to
achieve a transfer of skills from the cognitive collaboration where the person
is becoming more autonomous with time.
8. They afford a simple, powerful formalism of
thinking:
Using GAI applications as TFC means to engage in
complex activities that promote deep thinking when doing tasks with them and
not just base them on a stimulus-response education, or understanding them as
intelligent agents (with agency). This collaboration is not only based on
augmentation of action, but also on the reorganization of the activity. This is
the case of the study by Nguyen et al. (2024) on the use of a GAI writing tool
for PhD students, where iterations and interactions with the tool showed a
better performance in their writing compared with those who just used the tool
as a source of information. Also, a nursing education study (Simms, 2024) showed
that students could reflect on their questioning, the obtained responses and on
their decision-making in problem-solving (these are executive functions). So,
it contributed to a constructivist meaning making process.
9. They are easily learnable:
GAI applications have a manageable intrinsic cognitive
load that affects learning positively. In fact, when using them in a learning
task, they should help increase the germane cognitive load to improve the
process of acquiring new knowledge in the long-term memory. These tools follow
a familiar interface in the form of a chatbox. There
is not a cognitive load that affects its adoption as a TFC, just the danger of
understanding it as a reliable intelligent agent to learn from. AI literacy
should be added in Teaching Digital Competence frameworks such as Digcompedu (Punie & Redecker, 2017) so that both
educators and students are able to leverage the benefits of GAI applications
and to avoid their shortcomings.
6. Conclusions
We have offered a theoretical framework that places
GAI applications under the perspective of TFC in education. TFC are not a specific
kind of technology but a concept or a metaphor about how to integrate
technology in teaching and learning processes to empower constructivist and/or
socio-constructivist learning, where students use them as tools aimed at learning
with them, establishing a cognitive partnership, and where they have the
main role and agency, not the tool. This is the opposite view of learning from
GAI applications.
Sociocultural theory and activity theory work as an
overarching theoretical context to understand digital tools as media for
distributed activity between people, context, and tools. There is a
distribution of cognition between the person and the TFC so that the person can
go further in a joint system of human(s)-tool(s). Without it, the task could be
hard or even impossible to attain.
GAI applications meet the practical and pedagogical
requirements identified by David Jonassen (1996) to work as TFC. Thus, we have
updated Jonassen's earlier classification of "mindtools" adding these
specific AI tools. They stand out as TFC for critical thinking based on
prompting.
To become TFC, they must be set with a purpose to
reach an objective, some form of motivation for the learning activity. If we
want to leverage their potential in education, we must understand how TFC’s
affordances can promote learning. The TFC perspective redirects educational
practice from the individual without tools, or from the individual using
technology that acts as an artificial tutor (the traditional teacher-centered view of education), to the recognition of the
cognitive collaboration between students and TFC. It is not that we have the
possibility to include their use in education, but we must encourage it for a
future of cognitive partnerships that will equip students with skills (that
will not be person-solo based) that they are going to encounter in their future
professional life (DeFalco & Sinatra, 2019; IFTF, 2017; Perkins, 1993).
Digital tools should only be used for skill mastery
and not for deskilling students (Salomon 1993). TFC should help students think,
not help take over their cognition by just off-loading it, or by doing the whole
job. They should not perform students' executive functions (e.g.,
decision-making) but facilitate deeper thinking leveraging their epistemic
actions (Kim & Reeves, 2007). It is fundamental that when used in
educational settings, teachers design activities that integrate them but
assuring that students will use them to review their inputs and outputs, to
refine prompts and, eventually, improve their critical thinking skills, because
the responses we obtain from GAI applications may be opaque (Bearman & Ajjawi, 2023).
We are still in the early stages of integrating GAI
applications in teaching and learning and of knowing their affordances for
distributed cognition. We need more research to shed light upon these possible
distributed cognitions, to elicit the use of their epistemic actions.
References
Akata, Z., Balliet, D., de Rijke, M., Dignum, F., Dignum,
V., Eiben, G., Fokkens, A., Grossi, D., Hindricks, K., Hoos, H., Hung, H., Jonker, C., Monz, C., Neerincx, M., Oliehoek, F., Prakken, H., Schlobach, S., van
der Gaag, L., ..., Welling, M. (2020). A Research Agenda for Hybrid Intelligence: Augmenting
Human Intellect with Collaborative, Adaptive, Responsible, and Explainable
Artificial Intelligence. Computer,
53(8), pp.18-28, http://doi.org/10.1109/MC.2020.2996587
Alier,
M., García-Peñalvo, F.J., & Camba, J.D. (2024). Generative Artificial Intelligence in Education: From
Deceptive to Disruptive. International Journal of Interactive Multimedia and
Artificial Intelligence, 8(5), pp. 5-14. https://doi.org/10.9781/ijimai.2024.02.011
Bakhurst, D. (2009). Reflections on activity theory. Educational
Review, 61(2), 197–210. https://doi.org/10.1080/00131910902846916
Bearman, M.,
& Ajjawi, R. (2023). Learning to work with the
black box: Pedagogy for a world with artificial intelligence. British
Journal of Educational Technology, 54, 1160–1173. https://doi.org/10.1111/bjet.13337
Cain, W.
(2024). Prompting Change: Exploring Prompt Engineering in Large Language Model
AI and Its Potential to Transform Education. TechTrends,
68, 47–57. https://doi.org/10.1007/s11528-023-00896-0
Chomsky, N.,
Roberts, I., & Watumull, J. (2023, March 8). The
false promise of ChatGPT. The New York Times. https://shorturl.at/gvmOw
Clark, A.,
& Chalmers, D. (1998). The extended mind. Analysis, 58(1), 7–19. https://shorturl.at/wj9RI
Cole, M.
& Engeström, Y. (1993). A cultural-historical
approach to distributed cognition. In G. Salomon (ed.), Distributed
cognitions. psychological and educational considerations (pp. 1-46).
Cambridge University Press.
Crompton,
H., & Burke, D. (2024). The Educational Affordances and Challenges of
ChatGPT: State of the Field. TechTrends 68,
380–392. https://doi.org/10.1007/s11528-024-00939-0
Dawson, S., Joksimovic, S., Mills, C., Gašević,
D. & Siemens, G. (2023), Advancing theory in the age of artificial
intelligence. British Journal of Educational Technology, 54, 1051-1056. https://doi.org/10.1111/bjet.13343
DeFalco,
J.A. & Sinatra, A.M. (2019). Adaptive Instructional Systems: The Evolution
of Hybrid Cognitive Tools and Tutoring Systems. In R. Sottilare,
J. Schwarz (eds) Adaptive Instructional Systems. HCII 2019. Lecture Notes in
Computer Science, vol. 11597. Springer, Cham. https://doi.org/10.1007/978-3-030-22341-0_5
Derry,
S. J., & LaJoie, S. P. (1993). A middle camp for (un)intelligent instructional
computing: An introduction. In S. P. Lajoie & S. J. Derry (Eds.), Computers as cognitive tools (pp. 1-14).
Lawrence Erlbaum Associates. (pp.2-4).
Dubé, A. K.,
& McEwen, R. N. (2017). Abilities and affordances: factors influencing
successful child–tablet communication. Education Tech Research Dev. 65, 889–908 https://doi.org/10.1007/s11423-016-9493-y
Eager, B.,
& Brunton, R. (2023). Prompting Higher Education Towards AI-Augmented
Teaching and Learning Practice. Journal of University Teaching &
Learning Practice, 20(5). https://doi.org/10.53761/1.20.5.02
Engeström Y. & Sannino, A. (2021) From mediated actions to
heterogenous coalitions: four generations of activity-theoretical studies of
work and learning, Mind, Culture, and Activity, 28(1), 4-23, http://doi.org/10.1080/10749039.2020.1806328
Garcia
Brustenga, G., Fuertes-Alpiste, M. &
Molas-Castells, N. (2018). Briefing
paper: chatbots in education. eLearning
Innovation Center, Universitat Oberta de Catalunya. https:/doi.org/10.7238/elc.chatbots.2018
Gibson, J.
J. (1979). The ecological approach to visual perception. Houghton Mifflin Company.
Gibson, E.
J. (1989, July). Learning to perceive or perceiving to learn? Paper presented
to the International Society for Ecological Psychology, Oxford.
Holstein,
K., Aleven, V., & Rummel, N. (2020). A Conceptual
Framework for Human–AI Hybrid Adaptivity in Education. In: I. Bittencourt, M. Cukurova, K. Muldner, R. Luckin, E. Millán. (eds) Artificial Intelligence in
Education. AIED 2020. Lecture Notes in Computer Science, vol 12163.
Springer, Cham. https://doi.org/10.1007/978-3-030-52237-7_20
IFTF
[Institute for the Future]. (2017). The next era of human / machine partnerships. Emerging
technologies' impact on society and work 2030. Institute for the Future & Dell Technologies. https://shorturl.at/4DJUD
Iiyoshi, T., Hannafin, M. J., & Wang, F. (2005).
Cognitive tools and student-centered learning:
rethinking tools, functions and applications, Educational Media
International, 42(4), 281-296. https://doi.org/10.1080/09523980500161346
Järvelä, S.,
Nguyen, A., & Hadwin, A. (2023). Human and artificial intelligence
collaboration for socially shared regulation in learning. British Journal of
Educational Technology, 54, 1057–1076. https://doi.org/10.1111/bjet.13325
Jonassen, D.
H. (1996). Computers in
the classroom. Mindtools for critical thinking. Prentice Hall.
Jonassen, D.
H., Carr, C. Yueh, H. P., (1998). Computers as mindtools for engaging learners
in critical thinking. TechTrends, 43(2),
32-35. https://doi.org/10.1007/BF02818172
Kanuka, H.
(2008). Understanding e-learning technologies-in-practice through
philosophies-in- practice. In T. Anderson (Ed.), The theory and practice of
online learning (2nd ed., pp. 91–118). Athabasca University Press.
Kim, B.,
& Reeves, T. C., (2007). Reframing research on learning with technology: in
search of the meaning of cognitive tools. Instructional Science, 35,
207-256. https://doi.org/10.1007/s11251-006-9005-2
Kirschner,
P. A., & Erkens, G. (2006). Cognitive Tools and Mindtools for Collaborative
Learning. Journal of Educational Computing Research, 35(2),
199-209. https://doi.org/10.2190/R783-230M-0052-G843
Kirsh, D. &
Maglio, P. (1994). On Distinguishing Epistemic from Pragmatic Action. Cognitive
Science, 18, 513-549. https://doi.org/10.1207/s15516709cog1804_1
Knoth, N., Tolzin, A., Janson, A. & Leimeister, J. M. (2024). AI
literacy and its implications for prompt engineering strategies. Computers
and Education: Artificial Intelligence, 6, https://doi.org/10.1016/j.caeai.2024.100225
Kozulin,
A. (2000). Instrumentos psicológicos. La educación desde una perspectiva
sociocultural. Paidós.
Lave, J.
(1988). Cognition in
practice. Cambridge
University Press.
Leontiev, A. N. (1978). Activity,
consciousness, and personality.
Prentice-Hall.
Miao,
F., & Holmes, W. [UNESCO] (2023). Guidance for generative AI in education and research. UNESCO
Mishra, P.,
Oster, N. & Henriksen, D. (2024). Generative AI, Teacher Knowledge and
Educational Research: Bridging Short- and Long-Term Perspectives. TechTrends 68, 205–210. https://doi.org/10.1007/s11528-024-00938-1
Nguyen, A.,
Hong, Y., Dang, B., & Huang, X. (2024). Human-AI collaboration patterns in
AI-assisted academic writing. Studies in Higher Education, 1–18. https://doi.org/10.1080/03075079.2024.232359
Norman, D.
A. (1999). Affordance, conventions, and design. Interactions, 6(3),
38–43.
Open AI
(2024, May 10). Open AI. About. https://openai.com/about/
Papert, S. (1982). Mindstorms: children, computers and powerful ideas. Harvester Press.
Pea, R. D.
(1985). Beyond amplification: Using the computer to reorganize mental
functioning. Educational Psychologist, 20(4), 167-182. https://doi.org/10.1207/s15326985ep2004_2
Pea, R. D.
(1993). Practices of distributed intelligence and designs for education. In G.
Salomon (ed.), Distributed cognitions. Psychological and educational
considerations (pp. 47-87). Cambridge University Press.
Perkins, D.
N. (1993). Person-plus: a distributed view of thinking and learning. In G.
Salomon (ed.), Distributed cognitions. Psychological and educational
considerations (pp. 88-110). Cambridge University Press.
Punie, Y.,
& Redecker, C (Eds.). (2017). European
framework for the digital competence of educators: DigCompedu. Publications Office of the European Union. https://doi.org/10.2760/178382
Rheingold,
H. (1985). Tools For
Thought: The History and Future of Mind-Expanding Technology. The MIT Press.
Sabzalieva E. & Valentini, A. [UNESCO] (2023). ChatGPT and artificial intelligence in higher
education: quick start guide. UNESCO.
Salomon, G.
(1993). No distribution without individuals' cognition: a dynamic interactional
view. In G. Salomon (ed.), Distributed cognitions. Psychological and
educational considerations (pp. 111-138). Cambridge University Press.
Salomon, G.
& Perkins, D. (2005). Do technologies make us smarter? Intellectual
amplification with, of and through technology. In R. Sternberg, R. and D.
Preiss (eds.), Intelligence and Technology: The Impact of Tools on the
Nature and Development of Human Abilities (pp. 71-86). Lawrence Erlbaum.
Salomon, G.,
Perkins, D. N., & Globertson, T. (1991). Partners
in cognition: Extending human intelligence with intelligent technologies. Educational
Researcher, 20(3), 2-9. https://doi.org/10.3102/0013189X0200030
Schellaert, W., Martínez-Plumed, F., Vold, K., Burden, J.,
Casares, P., A. M., Sheng Loe, B., Reichart, R., Ó hÉigeartaigh, S., Korhonen, A., & Hernández-Orallo, J.
(2023). Your Prompt is My Command: On Assessing the Human-Centred Generality of
Multimodal Models. Journal of Artificial Intelligence Research, 77,
377-394. https://doi.org/10.1613/jair.1.14157
Scribner, S.
& Cole, M. (1981). The
psychology of literacy. Harvard
University Press.
Sharples, M.
(2023). Towards social generative AI for education: theory, practices and
ethics. Learning: Research and Practice, 9(2), 159–167. https://doi.org/10.1080/23735082.2023.2261131
Simms,
Rachel Cox DNP, RN, FNP-BC. (2024). Work With ChatGPT, Not Against: 3 Teaching
Strategies That Harness the Power of Artificial Intelligence. Nurse Educator,
49(3), 158-161. http://doi.org/10.1097/NNE.0000000000001634
Vygotsky, L.
S. (1978). Mind in
society. Harvard
University Press.
Walter, Y.
(2024). Embracing the future of Artificial Intelligence in the classroom: the
relevance of AI literacy, prompt engineering, and critical thinking in modern
education. International Journal of Educational Technology in Higher Education,
21(15). https://doi.org/10.1186/s41239-024-00448-3
Wertsch,
J. V. (1995). Vygotsky y la formación social de la mente. Paidós.