IEEE Robotics & Automation Magazine - June 2016 - 26

views. The teacher's role during the discussion should
consist mostly of asking children to clarify their own
views ("What do you mean by ...?"), justify them ("Why
do you think that ...?"),
find a way of justifying
them ("What could we
Robots present more
do to find out if you are
right?"), draw implicaadvantages than traditional tions or predictions
from specific assertions
objects of scientific
("If you are right, then
what should happen?"),
investigation as tools for
and identify similarities,
differences, and contrafostering the development
dictions among views
proposed by different
of scientific reasoning
children ("So do you
agree with B or not?").
abilities in young children.
Children's questions are
put back to the group
rather than answered by
the teacher. This dialogical approach, akin to the socalled coupled inquiry [9] method combining guided and
open teaching methodologies [10], encourages children
to be independent in reasoning about their observations.
This by no means implies that the teacher is only a passive participant. On the contrary, he or she must actively
orient the discussion along paths that are functional to the
GoS learning objectives and goal. However, this role must
be fulfilled by posing questions rather than by transmitting expert information. Examples are provided in the
"Results and Discussion" section below.
The Robot
Choice of the right robot (in terms of its physical structure and behavioral mechanism) is crucial to the success
of GoS. Ideally, the robot's behavior should be eccentric

Figure 1. A LEGO Mindstorms robot used in a GoS activity.

IEEE ROBOTICS & AUTOMATION MAGAZINE

june 2016

enough to stimulate children's interest without being too
difficult to explain. For example, in a GoS session with a
class of second-graders we used a LEGO Mindstorms roving vehicle with two front proximity sensors, oriented
slightly rightward and leftward, and a sound sensor (Figure 1). A simple Braitenberg-like algorithm [4], implemented using the BricxCC environment (bricxcc.
sourceforge.net), produced obstacle avoidance plus peculiar reactions to sounds.
The Activity
The GoS sessions conducted to date have been structured
as follows. We start the game by simply showing the robot
to the children while it is turned off and asking them to
describe it. The goal of this descriptive phase is to allow
the students to familiarize themselves with the robot's
physical structure and gather information to be used in the
subsequent explanatory stage. Next, the robot is turned on
and left to wander about an experimental arena. We have
experimented with two different methods of proceeding
from this point. The first is to ask children, "What is the
robot doing?," and then, after listening to their observations about its behavior, to advance explanation requests
of the kind "Why is the robot doing that?" The second
method is to ask the children themselves what they
would like to find out about the robot. The teacher takes
note of the research questions proposed by the children,
which are typically 1) questions about the robot's abilities ("Can it dance?"), 2) questions regarding the functional role of particular components ("What is this block
for?"), 3) explanation requests ("Why does it sometimes
go faster?"), and 4) questions about the nature of the
control mechanism ("Does it move randomly or not?").
These are only two of many possible ways of launching the explanatory phase of GoS. From this point
onward, the game involves collaboratively addressing the
questions raised by the children or the explanatory
requests made by the teacher. The teacher's justification
requests (see the "Teaching Style" section) typically stimulate the children to autonomously formulate and carry
out experiments and to discuss the suitability of the proposed experiments. For both activities, especially the latter, the class may be divided into groups. For example,
after the class has observed the robot in motion, students
may be assigned to small groups, each of which is asked
to produce a list of possible explanations for particular
aspects of the robot's behavior. Clearly, the structure just
outlined may be customized to suit the characteristics of
different classes.
Explaining Robotic Behaviors
The goal of GoS is to produce good explanations for the
robot's behavior. However, there is no clear consensus among
philosophers of science as to what constitutes a good explanation-in this case, for the behavior of a robot. Several
competing theories are currently under debate [11]. (Note

http://www.sourceforge.net

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2016