Sunteți pe pagina 1din 5

More Emotional Students Make For Better Teachers?

Remember those days in grade school when you had that monotonous algebra teacher who just
droned on and on and on? Now remember that physics teacher everyone loved who was always
so enthusiastic and made the classroom come alive? Looking back, you feel like you learned so
much more in physics than algebra! Its all Mr. Algebra Teachers fault! you tell yourself
But not quite! According to the findings from a 2011 social robotics study at Yale University, you
might be partly to blame for his poor teaching! From this study, researchers discovered that the
quality of human teaching improves upon receiving appropriate, emotional feedback from robotstudents. Granted, youre not a robot-student, but if we extend the implications of this study from
the realm of human-robot interaction to human-human interactions, perhaps emotional feedback
is a critical component for better teaching in the classroom setting as well.
Upon beginning their study, the researchers were interested in testing if the emotional content of
a robots speech affects how people teach it. They hypothesized that a robot who expressed
appropriate, highly emotional responses would elicit better human teaching than a robot who
expressed no emotion in its feedback. They also hypothesized that a robot who expressed
inappropriate, but still highly emotional responses would not elicit better human teaching. Sixtytwo human participants from the greater New Haven area were asked to teach a fuzzy yellow
robot, Keepon, to dance a total of five different dance sequences. They were asked to stand on a
Wii Fit board and to demonstrate moves based on instructions displayed on a screen behind
Keepon. As the subjects moved their bodies, Keepon appeared to be imitating their movements.
Then, after each dance sequence, a big screen revealed Keepons performance accuracy score for
both the teacher and student-robot to see.
Now at this deciding moment, Keepon would utter a response to his score. The words spoken
and emotional content of the response differed based on conditional group. In the control group,
Keepon uttered an unemotional, apathetic response, such as we did okay or oh, thatll do. In
one experimental group, Keepon uttered a highly emotional response that was appropriate for its

score, such as Ooh, check that out, we did great! for a very high score and Ugh, man, this is
hopeless! for a very low score. In the second experimental group, Keepon uttered a highly
emotional response that was inappropriate for its score, such as Oh no, that was terrible! for a
really high score or Look at that! That is an awesome score! for a really low score.
In all three conditional groups, each of Keepons responses was prerecorded and came from the
same female voice. Furthermore, the accuracy scores that Keepon received after each dance
sequence was predetermined, unbeknownst to the participants. In other words, the participants
did not know that regardless of how well or how poorly he/she actually demonstrated the dance,
the same already-established score was to be displayed after each demonstration. Furthermore,
this numerical score following each specific demonstration was kept constant across the three
conditional groups. For instance, whether the participant was in the control group or one of the
experimental groups, if he/she chose to teach Keepon the third dance sequence for a sixth time,
Keepons score was preset at 94%. Other factors that were kept constant included the songs
accompanying each dance sequence, the instructions each participant received before beginning,
and the position of the participant relative to the robot. The only variable that was changed, the
independent variable, was Keepons response after seeing its score. Within each of the three
conditional groups, Keepon produced the same feedback (same words and same emotional
content) for all participants.
Upon hearing Kepons feedback, the human teacher was then given a choice to either teach the
same dance sequence againthereby giving Keepon a chance to improveor to move on to the
next dance sequence. The maximum number of times the subjects were allowed to teach Keepon
was ten per dance sequence. The actual number of times a subject chose to teach the same dance
sequence was then counted and used as a proxy for quality of teaching. An assumption of this
proxy is that quantity implies quality. This assumption was reasonable for this study since
teaching the same dance more times indicated greater patience on the teachers part (and patience
is a crucial component of good teaching).
Meanwhile, the accuracy of the subjects own demonstrations was secretly recorded
(unbeknownst to them, the Wii Fit board was programmed to detect any mismatches in the

participants movements with the instructions displayed on the screen). These human scores were
then used as another proxy for quality of teaching. Here, the researchers assumed that all
participants knew how to use a Wii Fit board properly such that the recorded accuracies indeed
reflected their actual movements. Again, the assumption here was reasonable because the subject
pool was made up of mostly undergraduates from Yalewho have likely been exposed to Wii,
DDR, or other gaming equipment of similar nature.
At the end of the study, the researchers used a statistical T-tests to compare the mean accuracy of
each participants demonstrations across the three conditional groups. They then used another Ttest to compare the mean number of demonstrations per dance sequence across the conditional
groups. Results agreed with their original hypothesis: participants in the appropriate emotional
response group demonstrated the dances significantly more accurately (p<0.001) and more
frequently (p<0.001) than those in the control group; no statistically significant difference was
detected in either teaching quantity or accuracy between the inappropriate emotional response
condition and the control condition.
Specifically, subjects in the appropriate emotional response condition demonstrated each dance
an average of 5.9 times (standard deviation= 2.3) at an average accuracy of 89% (standard
deviation = 12%); subjects in the inappropriate emotional response condition demonstrated each
dance an average of 4.1 times (standard deviation = 1.5) at an average accuracy of 80% (standard
deviation = 15%); and subjects in the apathetic response condition (the control group)
demonstrated each dance an average of 3.8 times (standard deviation = 1.0) at an average
accuracy of 81% (standard deviation = 15%). These results led the researchers to conclude that
appropriate emotional feedback from robots elicits better human teaching.
Now, how did the researchers reach such a conclusion? How did they get from twelve separate
numbers (six mean values and six standard deviation values) to the phrase statistical
significance? In order to understand the statistics they used, we must demystify the method of
T-testing. A T-test basically calculates whether the means of two data sets (one from an
experimental group and another from the control group) are significantly different. First, a null
hypothesisa hypothesis that there is no meaningful difference in the two groups data setsis

formed. Then, a T-value, or signal-to-noise ratio, is calculated from dividing the difference
between the two means by the standard error of such difference. The larger the magnitude of the
T-value, the more dissimilar the two data sets are from each other. Finally, to determine
significance, statisticians must calculate the probability of obtaining a T-value as large or as
small as the one actually obtained if the null hypothesis was indeed true. This probability is
called the p-value. When p-values are less than 0.05, statisticians say they reject the null
hypothesis, meaning the two original data sets have significantly different means. On the other
hand, when p-values are greater than 0.05, statisticians say they fail to reject the null
hypothesis, meaning no significant difference can be deduced between the two sets of data.
In this study, all T-tests that compared the control groups data to the appropriate emotional
response groups data produced p-values less than 0.001. Such tiny p-values indicated very high
levels of significance, meaning we should trust these Yale researchers conclusion that the
appropriately emotional Keepon elicited more accurate and more frequent human teaching.
Intuitively, their conclusion certainly makes sense too!
On the other hand, all T-tests that compared the control groups data to the inappropriate
emotional response groups data produced p-values greater than 0.05. These larger p-values
indicated that inappropriate feedback from Keepon, even if highly emotional, failed to improve
the accuracy and frequency of human teaching. Here, again, we should trust the Yale researchers
conclusion because it makes sense: who would want to keep teaching a robot that considers a 94
terrible?
But no matter how credible you consider this studys findings, at the end of the day, you might
still be wondering, Why does all this matter? If Im not a crazy robot scientist, why should I
care? If you find yourself asking these questions, its because you havent realized the
implications of this study in the field of education! Think back to that boring algebra teacher you
had whose voice made the whole class fall asleep. Did the class first become bored because he
was a boring lecturer? Or did he become a boring lecturer after receiving consistently apathetic
feedback from his students? What if you had tried harder to pay attention and pretended like you
were interested in algebra? Could your attentive and highly emotional responses have elicited

better teaching from him? According to this human-robot interactive study, the answer to that last
question is very likely, Yes!
With all the recent talk about Race to the Top and improving public school curricula, what if we
approached educational reform via a novel method: by trying to increase the emotional feedback
students provide their teachers? Of course, it would be very hard to induce nationwide change in
the behaviors and attitudes of students, but thinking about todays educational problems in this
new light is certainly a start! And once we start, we can Keepon dancin!

S-ar putea să vă placă și