To develop arti cial intelligence (AI) that educators can adopt in general educational environments, we are examining the potential role of AI in the socio-cultural aspects of learning in human development. In this position paper, we propose an experimental design, Creative Peer System, where humans and machines learn from each other in a multimodal learning environment and develop original artifacts. The research is in the early stage, where we are actively developing new types of empirical studies. We will present the methodological and theoretical frameworks and a design proposal that can elicit constructive feedback toward further re nement and implementation of the experiment.
Creativity is considered an essential learning outcome for the 21st century that
values problem solving and communication. Students need a broad range of
knowledge, like understanding how things are made and accomplished [
Many types of research that examine arti cial intelligence (AI) as an
educational technology re ect on this interest in academic knowledge. AI and
education research has many exciting developments that can increase personalization,
reduce administrative labor, support assessments, and increase access to
education [
To explore the role of AI in creative capacity building, especially in early childhood education, we have set the following design research questions: { How can AI be meaningfully incorporated into a socio-cultural learning environment?
To study the phenomenon, we focus on the following sub-research questions: { How can AI enhance the creative capacity of the students? { How can AI themselves be considered creative?
This article proposes an experimental design, Creative Peer System, where humans and machines learn from each other in a multimodal learning environment and develop original artifacts. The research is in the early stage of development, where we are considering various approaches and methods for new empirical studies. This article is more a position paper than a research paper presenting study results. In the following sections, we will brie y present the methodological and theoretical framework in consideration and an early design proposal that results from thought experiments carried out while studying the design context of AI in education. We conclude by discussing the potential impact for a wider researcher and educator communities by opening up the scope of AI in education toward new types of data sources, systems, and environments. 2
We rely on research-based design and design-based research methodologies
interested in design artifacts and pedagogical interventions in educational settings
and study their e ects in a real educational context [
With the experiment, we aim to examine the role of AI critically in a
learning context. As a theoretical framework, we use the cultural-historical activity
theory [
Often AI as a learning technology is regarded as a tool within a learning
environment [
We assume that as a learner is expanding one's capacity in a learning
context, AI may become a peer that grows with learners in the zone of proximal
development as described by Vygotsky. The learner grows when interacting with
a more knowledgeable or skillful peer [
ZONE OF PROXIMAL DEVELOPMENT STUDENT
AI AS A PEER
Learners cannot create yet Learners can create with peers Learners can create without peers
In assessing the creative capacity of the learners and machine, the experiment
will have uninformed assessment criteria between human and machine
learners. The experiment will develop assessment criteria developed in computational
creativity: combinatorial creativity, evolutionary creativity, and transformative
creativity [
In developing AI that exhibits creative behaviors, the experiment will use the
generative adversarial network, reinforcement learning, and embodied learning
through robotics and virtual simulation [
The Creative Peer System is a socio-cultural learning environment where humans and machines learn from each other and produce a new form of creative artifacts. Below is an experimental design, a design concept, that can test the educational impact of the system in the context of early childhood education.
The experiment has three types of participants in the system. { Students: ve to ten kindergarten students at the age between four and seven. { Teachers: one or two early childhood educators and at least one machine evaluator supervised by programmers. { Machines: ve to ten robots with learning and generative capacities (AI). The system operates in a physical environment that enables peer-to-peer learning, making, and re ection to replicate the common practices of socio-cultural learning. The example system is a typical kindergarten art class in an open space that allows peer-to-peer observation and access to the same materials. The machines use multi-sensory techniques like camera vision and audio detection to observe their peers and assess the sense of presence in the environment.
ASSESSMENT REFLECTION & EVALUATION
CREATIVE PEER SYSTEM
ITERATIVE CYCLE
GALLERY OF ITERATIVE ARTIFACTS
INSTRUCTION TEACHER
STUDENT
MACHINE
MAKING & OBSERVATION
The system will have interactive cycles of activities involving the students, machines, and teachers. Each iteration cycle will go as follows: { Instruction: Students and machines get situated side by side in an open space, receiving the same making materials and instructions. { Making and Observation: Students and machines create their artifacts while observing others' progress freely. { Re ection and Evaluation: Once the artifact creation is completed, the students and machine share their works in a shared space, observe, and re ect. { Assessment: Teachers assess the artifacts considering three aspects of creativity (combinatory, exploratory, and transformative) using visual perception, interpretation, peer-to-peer discussion, and algorithmic categorization.
The system will repeat multiple iterations while documenting the interactions, artifacts, and assessments. After the series of iterations, the researcher can assess the creative capacity development of the group by tracking evolutionary behaviors such as stylistic transfers, di usion, development of a new form of art, and new norm establishment.
In order to evaluate the educational impact of the system, comparable systems can be designed and tested with di erent groups of students. While the making activities and physical space stay the same, multiple variations can be tested, such as machines functioning in a di erent role, for example, tutor, tool, or absent, in a socio-cultural learning environment.
By evaluating the students' and machines' behavior and artifacts, the experiment can examine in what capacity AI can be incorporated into a socio-cultural learning environment. The experiment can reveal various ways AI in uences the students' creative capacity and if AI itself can be considered creative by developing something original and transformative to its peers. 4
Many topics need further consideration in this experimental design. While both human educators and machines will be used for evaluating the artifacts' creative development, there need to be systematic ways to measure both human creative capacity and artifacts' artistic evolution.
Many existing machine learning techniques require a high volume of data and numerous iterations to train and generate artifacts that are noticeably di erent from their previous outputs. Given these current technical constraints, there would need to be pre-training with data outside of the experiment, analogous to the students coming in with previous life experience. In addition, computational design and processing need to be up to the students' making and observation speed in real-time.
While the experiment seeks to address the machine's creative capacity in the socio-cultural learning context, it is debatable that this argument is limited only to the scope of the speci c environment and learning activities. However, a Creative Peer System can be used to further experiment with the dynamics of creative capacity building in di erent environments, age groups, and objectives. These tests can contribute to a broader understanding of the machine and human creativity.
In this position paper, we propose that AI and learning science researchers look at the socio-cultural environment's role in human learning, especially in learning creativity. We propose an approach where AI is primarily considered a creative peer of the learners to develop its creativity and, in turn, help learners develop their creativity. We call this the Creative Peer System, a socio-cultural learning environment where humans and machines learn from each other and produce a new form of creative artifacts. As the research is still in its early stages, this paper can elicit constructive feedback toward further re nement and implementation of the experiment. We hope this work can positively contribute to expanding AI works as a multimodal educational technology in a general educational environment.