include three additional families of pedagogical
agents that serve functions very different from, but
also complementary to, the classic pedagogical
agents we proposed: teachable agents, learning companions, and role-playing agents. While classic pedagogical agents are fundamentally teachers, teachable agents (Biswas et al. 2005) are fundamentally
students. Learners interact with teachable agents by
“teaching” them: learners introduce new concepts
and explain the relationships between concepts to
teachable agents, usually through a GUI that supports concept map creation. For example, in Betty’s
Brain, students teach teachable agent Betty (figure 3)
about science topics, and they monitor Betty’s learning as she takes quizzes and responds to questions
(Biswas et al. 2005). Interactions with teachable
agents elicit the self-explanation effect (Chi et al.
1989), which promotes deeper learning through
knowledge integration and mental model refinement. Teachable agents also appear to elicit the protégé effect (Chase et al. 2009), which is that learners
seem to feel responsible for their teachable agents
and to exert more effort to learn for their agents than
when learning alone without an agent.
While classic pedagogical agents teach and teachable agents learn, learning companions, such as
those depicted in figure 4 (Karumbaiah et al. 2017),
act as peers to learners (Chou, Chan, and Lin 2003).
Rather than acting didactically as an authority figure
or requesting assistance as a student, learning companions act as knowledgeable peers, and play an
important social role in learning (Kim and Baylor
2006). Learning companions can serve a key motivational function, which can be moderated by many
factors such as gender. For example, a recent study
found that a learning companion deeply integrated
into the narrative of a game-based learning environment produced experiences that were significantly
more engaging for girls than for boys compared to a
learning companion without the same backstory and
personality, even holding task support constant and
controlling for learners’ prior knowledge and video
game experience (Pezzullo et al. 2017). In addition to
“pure” learning companions, some learning companions provide both a cognitive and social role.
Cai et al. (2014) have experimented with “
tria-logues” where learners interact with a pair of agents
that combine these agent roles. One agent plays the
role of expert agent and another the role of fellow
student. Including role-playing agents makes it possible to vary the interaction strategy depending upon
the learner’s level of ability. Low-ability learners can
learn vicariously by watching the expert agent teach
the student agent. Medium-ability learners can
engage the expert agents in tutorial dialogue. High-ability learners can teach the student agent.
Agents increasingly play roles as participants with-
in interactive learning scenarios and games. In our
2000 paper, we suggested that agents could act as vir-
tual teammates, which has proven to be hugely
important for learning foreign languages, cross-cul-
tural skills, and interpersonal skills more generally.
Applications first appeared in military training (John-
son 2010) and are now being used in healthcare
(Taglieri et al. 2017), education, and corporate train-
ing. Such agent-based scenarios help learners gain
skills that readily transfer to the real world by
enabling them to learn in safe environments where
they can practice and make mistakes with impunity,
in contrast to live role-play exercises where learners
often feel that they are performing before an audi-
ence and are being judged.
Figure 5 shows a combination of agents used in
VCATs (Virtual Cultural Awareness Trainers) (John-
son et al. 2011), web-based cultural awareness cours-
es that have trained over 100,000 military service
members and are available for over 90 countries. A
Articles
Figure 2. Pedagogical Agent Kim in the Crystal
Island Game-Based Learning Environment.
Figure 3. Teachable Agent Betty in the
Betty’s Brain Learning Environment.
precipitation forms
reduces
condensation
vegetation
carbon dioxide
absorbs
produces
deforestation
vehicle used
+
––
–