The Wider Supportive Role of Social Robots in
the Classroom for Teachers
Paul Baxter1, Emily Ashurst2, James Kennedy1,
Emmanuel Senft1, S´everin Lemaignan1, and Tony Belpaeme1
1Centre for Robotics and Neural Systems
The Cognition Institute, Plymouth University, U.K.
2School of Nursing and Midwifery
Plymouth University, U.K.
Abstract. Robots are being increasingly used in schools by researchers
keen to assess how they may be used to facilitate learning and provide
support. Based on 15 school experiment visits at 9 diﬀerent schools in the
U.K., we outline our observations, speciﬁcally focusing on the broader
implications of robots in the classroom primarily from the perspective of
the teacher. We then outline the basis for future research considerations
in HRI, centred around the three themes of pedagogy, methodology, and
ethics. For further application of robotics to education, we suggest that
these three themes need to form a central part of continuing research.
Keywords: Education, Ethics, Methodology, Pedagogy, Social HRI,
There are increasing numbers of applications of social robots to the domain of
education, and in particular to deployment in classrooms. There are typically
two main, and often overlapping, goals for these eﬀorts. Firstly, they are intended
to augment teaching structures and provide supplementary support to children
by providing an alternative and/or personalised learning experience. Note that
these applications, quite rightly, do not seek to replace any human teaching staﬀ
or reduce human contact time, but are rather intended as supplementary to
existing pedagogical structures. Secondly, they seek to examine the attitudes of
teachers and students regarding robots in the classroom, and solicit their views
on how applications should be implemented and used.
Rather than ﬁt into one of these two essential goals, we rather in this paper
seek to take a step back and address some of the wider implications of robots
in the classroom from the perspective of the teachers themselves. We further
consider research methodology, beyond the application to the learning task that
is typically the focus of the work conducted.
We report on the experiences gained from 15 separate school visits to nine
diﬀerent schools around Plymouth (in the south-west of the U.K.) to run a num-
ber of experiments over the past three years. Our observations are qualitative
and based on extensive interactions during and after the experiments with teach-
ers. Despite this, a number of trends and facets have emerged that we believe
have consequences for the conduct of HRI experiments with children in schools,
but also more generally ‘in the wild’ – a diﬃcult task with many pitfalls .
Three outcomes from these observations are derived. Pedagogically, the way
the teachers use the robot in terms of its (social) presence needs to be properly
characterised from the perspective of teaching practice. Methodologically (from
the experimental point of view), the manner in which children anticipate their
interactions with the robot are altered by the changed behaviour of the teacher
(beyond the mere presence of the robot) needs to be accounted for. Ethically,
the way the robot is treated by the teacher with respect to the children and the
consequences of this needs to be considered.
2 Robots for Teachers
As noted above, much existing work is focused on having a robot support and
supplement existing teaching structures, by, perhaps, engaging children in addi-
tional one-on-one tuition. While space constraints prevent a complete overview
of the literature here, this is typically conducted from the perspective of the
individual children and/or of the robot itself. For example, Kanda et al provided
a seminal exploration of how robots interacted and formed relationships with
children over time , considering both the social dimension from the point of
view of the children, and the technical and behavioural facets of the robot used
to encourage this. Similar to this, the focus on improving robot competencies
to facilitate interaction with children has encompassed empathic behaviour ,
interactant-directed behavioural cues , etc.
The second strand of work is on teachers’ and children’s perceptions of robots
and how they may be useful. This has for example highlighted practical concerns
of robustness and equality of access, but also acknowledged the positive role of
the robot in maintaining the engagement of children . In the domain of
children with health problems, a social robot has been highlighted as a means of
integration at school (a novel experience to share that is not health-related) .
3 Teachers with Robots
Beyond the established goals described above, in this section we seek to expand
our observations of other factors involved in running experiments in real schools
that typically do not explicitly form part of the experimental considerations (or
appear in the subsequent publications). We ﬁrst provide some general observa-
tions, and then provide more speciﬁc observations from a recently conducted
3.1 General Observations
In the majority of our experiments (e.g. ), we set up the robot in a room
in the school that is not the children’s classroom, but one with which they are
familiar. In this way, we maintain the school environment while not disturbing
the ongoing lessons in another room. Children from our subject group (typically
a single class) are brought by the experimenter one by one to this experiment
room, the experiment is conducted, and the children are subsequently returned
to their classroom.
One consistent observation from our experiments in the schools was that the
teacher adapted their behaviour both in the presence of the robot and also given
just the prospect of the robot being present. This is not so surprising given that,
at the present time at least, robots are not a typical feature in U.K. schools.
A certain level of excitement at a novel experience in school time is thus to be
However, it was apparent that the teachers, without exception in our experi-
ence, would use the occasion to aid in their management of the classroom. There
were two general means of achieving this observed. Firstly, prior to the start of
the experiments, the teachers would emphasise to the children the importance
of getting the parents/guardians to complete the consent form - otherwise the
children would miss out. Secondly, during the experiment, the teachers would
explicitly refer to the robot when requesting that the children settle down to
work - a common phrase used was “if you don’t behave/be quiet, you won’t get
to play with the robot” (emphasis added3).
It must be acknowledged that the teachers would use similar tactics to en-
courage and remind the children to do various other normal school activities4.
However, in the case of the robot, this has a number of methodological conse-
quences. By emphasising that the children would miss out on the novel experi-
ence, the teachers are implicitly raising the expectation in the children that this
is a desirable experience. The children are therefore primed, with consequences
for subsequent interactions with the robot: it has indeed been shown that prior
expectation strongly shapes human-robot interactions, e.g. .
3.2 Long-term Embedded Study Observations
In one of our more recent studies, we dispensed with the presence of experi-
menters, and placed robots directly into classrooms without technical supervi-
sion: the teacher (and perhaps a teacher’s assistant) was the only supervising
adult present. In this continuous two-week study (publication in preparation),
the robot interacted with individual children during class time, in the classroom,
where the teacher was responsible for designating which child would be allowed
to interact with the robot next. Two classes were involved in the same school
3This use of the word ‘play’ seems to arise irrespective of the way in which the robot is
presented by the researchers. This could indicate that despite the insistence that we
are attempting to help the children learn, the robot does give rise to the expectations
associated with toys. Alternatively, it could just be the manner the teachers choose
to portray the robot in order to maximise the eﬀect with the children: play is fun,
work is not.
4Perhaps implicitly taking advantage of the children’s desire not to be left out, and
to remain included.
and in the same age-range, with separate teachers. In initial planning discussions
with the teachers, they enquired whether it would be possible to use the robot
to indicate to them when the classroom was too noisy to help them manage the
children, however this interfered with the planned experimental protocol and so
was subsequently not pursued. Based on observations from video recorded inter-
actions, and subsequent discussion with the teachers, this long-term embedded
situation provides additional insight into the altered dynamics in the classroom
when a (social) robot is present.
Firstly regarding the general attitude of the children over the two week pe-
riod. Both teachers involved said that while in the ﬁrst week the behaviour of
the robot caused distraction for the rest of the class (e.g. children turning to
look when the robot spoke), in the second week, this distraction eﬀect wore oﬀ.
This meant that the classes continued uninterrupted by the robot despite the
interactions taking place during normal lesson time. Conversely, both teachers
reported that their ability to use the robot as a ‘bribe’ (i.e. a means of motivat-
ing the children) was maintained over the experimental period. In other words,
while the distraction caused by the robot decreased, it still formed a suﬃciently
attractive lure to induce behaviour change.
Secondly, prior to this experiment, a great deal of eﬀort was put into designing
the content to be learned by the children. We chose to maintain a strong link
to the curriculum that the children were following, and also chose a topic that
would appear in the syllabus for the following year (thus a novel topic for the
children at the time of the experiment). This was done for two reasons. Firstly,
any positive learning outcomes would be directly useful for the children (and
thus also increasing ecological validity). Secondly, it was a means of making
the experiment relevant to the school, and reducing the potential (perceived)
distraction from the curriculum if a completely unrelated topic were chosen.
In all of the examples described above, it is clear that the novelty eﬀect provides
a signiﬁcant inﬂuence. As a means of increasing interest in a topic to be learned,
this eﬀect can be bootstrapped to improve child learning, e.g. . However, as
robots become more pervasive in these environments, there will naturally come
a point where this eﬀect of novelty can no longer be relied upon to increase
the interest levels suﬃciently to enhance engagement and/or learning. While
long-term studies have examined relationships between children and robots in
classroom settings , the equivalent research for learning outcomes has yet to
provide conclusive evidence that learning beneﬁts will remain, although there are
indications that at a least a moderate positive eﬀect will persist. Based on these
considerations, we can identify three themes around which our observations and
their implications can be structured: pedagogy, methodology, and ethics.
Pedagogy: Assuming that the goal is to enhance the classroom as a learning
environment with social robots, the wider question is how the robot would ﬁt
in terms of both the teaching style of the teacher and the content to be learned
by the children. Insight can be gained from research in the ﬁeld of Computer
Supported Cooperative Work/Learning, which has approached the problem from
the more general perspective of technology in the classroom. For example, the
concept of ‘classroom orchestration’  investigates how device appearance and
functionality aﬀords improved teacher-directed workﬂow. Taking advantage of
frameworks such as this will enable the role of the robot to be considered from
a global educational, rather than robot-centred technological, perspective.
Methodology: Given our observations, it would appear that the methodolog-
ical consequences need to be handled. If the children’s expectations are inﬂated
prior to the interaction with the robot by implicit suggestions from the teacher,
then a practical consequence could be a reduction in measurable diﬀerence be-
tween conditions. Consider that in many studies, a single manipulation distin-
guishes between conditions, and it is clear that an initial increase of excitement
or engagement in all subjects could be enough to reduce the signiﬁcance of any
diﬀerences between conditions – made already diﬃcult due the child propensity
to anthropomorphise . This could be an eﬀect that distinguishes research with
children from research with adults for instance. The means of handling this po-
tential confound could however be diﬃcult. One means could be to extend the
length of studies so as to get over this initial biasing eﬀect. Although this may
be practically diﬃcult, another beneﬁt of this solution is that begins to address
the questions in a more ecologically valid manner.
Ethics: A further consideration is how the manner in which the teacher treats
the robot impacts on the dynamics of the classroom, and the subsequent way
the robot is viewed by the children. This goes beyond the existing concerns of
the long-term eﬀect of children interacting with robots (e.g. in the extreme case
). One example given above was of the teacher requesting that the robot be
used to indicate if the classroom was too noisy. By using the robot in the capacity
of an ‘informant’ in this way, the children in the classroom would begin to see
it as an ‘other’ rather than as a peer (if the robot designers envisaged a peer
role for the robot). Similar scenarios can be constructed that follow similar lines
of argument: a mismatch between the roles apparently ascribed by the teacher
to the robot, and the role the robot takes on when interacting with a child. To
the extent that such discrepancies impact trust, for example, this becomes an
ethical issue that requires consideration.
In this paper we have outlined the three themes of pedagogy, methodology and
ethics as areas that are impacted by the presence of (social) robots in the class-
room, but which are not typically considered in research eﬀorts at the present
time. The example from methodology demonstrates what we contend are real
experimental issues that can be addressed if this perspective is taken. These ob-
servations are not intended to be read as criticisms of current practice (we have
been subject to the same eﬀects mentioned here), but rather as opportunities
for further research endeavours. What we advocate though is a wider consider-
ation of the role of the robot within the classroom beyond the actual learning
application that is typically the focus of the studies reported in the literature.
This work was supported by the EU FP7 ALIZ-E project (grant number 248116),
and the EU FP7 DREAM project (grant number 611391), www.dream2020.eu.
1. Baroni, I., Nalin, M., Baxter, P., Pozzi, C., Oleari, E., Sanna, A., Belpaeme, T.:
What a robotic companion could do for a diabetic child. In: RoMAN’14. pp. 936–
941. IEEE Press, Edinburgh, U.K. (Aug 2014)
2. Belpaeme, T., Baxter, P., Greeﬀ, J.D., Kennedy, J., Looije, R., Neerincx, M., Ba-
roni, I., Coti, M.: Child-Robot Interaction: Perspectives and Challenges. In: ICSR.
pp. 452–459. Springer, Bristol, U.K. (2013)
3. Dillenbourg, P.: Design for classroom orchestration. Computers and Education 69,
4. Fischer, K., Soto, B., Pantofaru, C., Takayama, L.: Initiating Interactions in Order
to Get Help: Eﬀects of Social Framing on Peoples Responses to Robots Requests
for Assistance. In: RoMAN’14. pp. 999–1005. IEEE Press, Edinburgh, U.K. (2014)
5. Kanda, T., Hirano, T., Eaton, D., Ishiguro, H.: Interactive Robots as Social Part-
ners and Peer Tutors for Children: A Field Trial. Human-Computer Interaction
19(1), 61–84 (Jun 2004)
6. Kanda, T., Sato, R., Saiwaki, N., Ishiguro, H.: A two-month ﬁeld trial in an elemen-
tary school for long-term human-robot interaction. IEEE Transactions on Robotics
23(5), 962–971 (2007)
7. Kennedy, J., Baxter, P., Belpaeme, T.: The Robot Who Tried Too Hard: Social
Behaviour of a Robot Tutor Can Negatively Aﬀect Child Learning. In: HRI’15. pp.
67–74. ACM Press, Portland, Oregon, USA (2015)
8. Leite, I., Castellano, G., Pereira, A., Martinho, C., Paiva, A.: Modelling Empathic
Behaviour in a Robotic Game Companion for Children : an Ethnographic Study
in Real-World Settings. In: HRI’12. pp. 367–374. ACM Press, Boston, MA, U.S.A.
9. Salter, T., Werry, I., Michaud, F.: Going into the Wild in Child-Robot Interaction
Studies. Intelligent Service Robotics 1(2), 93–108 (2007)
10. Schiefele, U.: Interest, Learning, and Motivation. Educational Psychologist 26(3-4),
11. Serholt, S., Barendregt, W., Leite, I., Hastie, H., Jones, A., Paiva, A., Vasalou,
A., Castellano, G.: Teachers’ views on the use of empathic robotic tutors in the
classroom. RoMAN’15 (September 2015), 955–960 (2014)
12. Sharkey, N., Sharkey, A.: The crying shame of robot nannies: An ethical appraisal.
Interaction Studies 11(2), 161–190 (2010)