ERIC EJ1114767: An Artificial Intelligence Tutor: A Supplementary Tool for Teaching and Practicing Braille

2016·0.41 MB·English

by  ERIC

CE Article
An Artificial Intelligence Tutor:
A Supplementary Tool for Teaching
and Practicing Braille
Tessa McCarthy, L. Penny Rosenblum, Benny G. Johnson,
Jeffrey Dittel, and Devin M. Kearns
Structured abstract: Introduction: This study evaluated the usability and effec-
tiveness of an artificial intelligence Braille Tutor designed to supplement the in-
struction of students with visual impairments as they learned to write braille
contractions. Methods: A mixed-methods design was used, which incorporated a
single-subject, adapted alternating treatments design as well as qualitative teacher
interviews and surveys. Results: Students seemed to reach 100% accuracy faster
when using Braille Tutor (average (cid:2) 7.0 sessions; range 1.0 to 12.0 sessions) than
when they did not (average 9.6 sessions with a teacher of students with visual
impairments;range3.0to16.0sessions).Also,studentswhousedBrailleTutormore
often tended to learn more contractions overall during the study (average (cid:2) 21.25;
range 13.0 to 30.0) than students who used it less (average 9.0; range 9.0 to 9.0).
Discussion: The first trend noted was that students in the teacher of students with
visual impairments plus Braille Tutor phase (hereafter, TVI(cid:3)Tutor) tended to learn
contractionsmorequickly.Asecondtrendsurfaced:ThestudentsintheTVI(cid:3)Tutor
phase tended to get more frequent reinforcement as opposed to students in the TVI
Onlyphase.Athirdtrendwasnoted:StudentsintheTVI(cid:3)Tutorphasesawaquicker
initial jump and tended to be more consistent in that initial jump. Although the
prototypeversionofBrailleTutorinthisstudyneedsfurtherdevelopmenttobroaden
its capabilities, some students found its use highly motivating. Implications for
practitioners: Although there is strong evidence that advanced technologies are not
suitable replacements for braille literacy instruction, technology can be used along
with quality instruction by a teacher of visually impaired students to enhance
proficiency in braille literacy.
B
raille is the primary literacy medium uals with visual impairments of working
for those who are blind. Braille literacy age found that the daily use of braille had
strongly correlates with better reading a positive impact on employment, salary,
habitsandinvolvementinpost-secondary and self-esteem (Bell & Mino, 2013).
education (Ryles, 1996). There is also a Approachestoteachingbeginningbraille
strong link between braille literacy and readers vary; however, a commonality
employment: A survey of 1,056 individ- among approaches is that learners need
309
©2016AFB,AllRightsReserved JournalofVisualImpairment&Blindness,September-October2016
CE Article
theopportunitytousebrailleanddevelop ininequityintheamountoftimestudents
their knowledge of braille contractions whoareblindspendinliteracyinstruction
(Swenson, 2016). The Alphabetic Braille compared with their sighted peers (Wall
and Contracted (ABC) Braille Study, the Emerson et al., 2009). Thus, a tool that
only longitudinal study of beginning can assist in supporting the literacy skill
braille readers, followed 38 readers from development of beginning braille readers
2002 to 2007. Although it was not clear andallowstudentstopracticebraillecon-
from the research if beginning braille in- tractions in the absence of teachers of
struction with contracted or uncontracted visually impaired students or other adults
braille increased the student’s later liter- whoknowbraillecouldproveinvaluable.
acy abilities, the researchers concluded Intelligent tutoring (that is, adaptive
that “all things being equal, the introduc- computer instruction) may help teachers
of visually impaired students provide
tion of contractions early in a student’s
their students with practice in developing
reading process is associated with higher
their braille skills at times when teachers
literacyperformancelaterinthestudent’s
are not present to provide reinforcement
literacy career” (Wall Emerson, Hol-
or answer questions. The third and fourth
brook, & D’Andrea, 2009, p. 622).
authors conducted a national survey of
Teachers of students with visual im-
teachers of visually impaired students to
pairments have many job roles, including
assess if there was a need for a tutoring
theresponsibilityforteachingtheexpanded
program that would provide reinforce-
corecurriculumintheareaofcompensatory
mentofbraillecontractionsbeinglearned
skills, a component of which includes the
by students. Responses from 68 teachers
useofbrailleforreadingandwriting.They
ofstudentswithvisualimpairments(84%
must also ensure that students have aca-
of whom were itinerant teachers) con-
demicsupportandareprovidedwithmate-
firmed the potential value of using intel-
rials in their literacy medium (Allman &
ligent tutoring software, with 90% and
Lewis, 2014; Griffin-Shirley, Koenig, &
88% rating the opportunity for additional
Layton, 2004; Wolffe et al., 2002); ensure
one-on-one tutoring in braille as having
other teaching responsibilities are met
moderate,high,orveryhighvaluetotheir
(Griffin-Shirleyetal.,2004;Wolffeetal.,
students working at school and at home,
2002); and manage large caseloads.
respectively (similar ratings were ob-
Griffin-Shirley and colleagues (2004) re-
tained,interestingly,forpotentialvalueto
ported that, on average, the teachers they
the teachers of visually impaired students
surveyed had 22 students on their case-
themselves).
loads, including two students who were
Despite the major strides achieved in
blind. These large caseloads often result
access technologies such as synthesized
speech (for example, screen readers),
brailleremainsanimportanttoolforread-
EARN CES ONLINE
ers with visual impairments. The exclu-
by answering questions on this article.
For more information, siveuseofaudiocanresultindeficiencies
visit: (cid:4)http://jvib.org/CEs(cid:5). in spelling and composition skills, as
pointed out by Foulke (1979) more than
310
JournalofVisualImpairment&Blindness,September-October2016 ©2016AFB,AllRightsReserved
CE Article
three decades ago. Others have argued either one-cell whole word contractions
that full reliance on audio is inconsis- representedbyasingleletterofthealpha-
tent with an operational definition of bet or short-form words, and (b) did not
literacy, which includes writing (Tuttle have additional significant intellectual
& Hatlen, 1996). Wittenstein and disabilities. The teachers had to have ac-
Pardee (1996) reported that 89% of cess to a computer with Internet, speech,
teachers of visually impaired students and a braille display.
agreed that speech technology should Recruitment occurred in the summer
be used as a supplement to braille, not and early fall of 2013. Advertisements
as a replacement. For sighted readers, were placed on electronic discussion
progress in technology has obviously groups in the field of visual impairment.
not replaced print—just the opposite: it Teachers who had prior experience in
has greatly simplified and empowered working with one of the authors were
access to the printed word. Similarly, contacted individually. Teachers who ex-
the use of an intelligent electronic pressed interest in the study and had a
braille tutor can be used to augment student or students who met the criteria
instruction by teachers of visually im- were sent consent packets for both them-
paired students and to promote in- selves and the children’s families.
creased levels of braille literacy. This
study sought to evaluate the impact of TEACHERTRAINING
such a braille tutor with the following Prior to the start of the study, a one-hour
questions: online training session was held. Partici-
pants viewed a demonstration of how to
• IsBrailleTutoreasytouse?Whatwould use Braille Tutor and its features, how to
increase its usability? administertheassessmentsusedintesting
• What are users’ experiences as they and intervention, and the basics of the
become oriented to Braille Tutor? studydesign.Annotatedversionsofthese
• What kinds of problems do users have presentations were provided to the teach-
in getting started with Braille Tutor? ers for reference. Throughout the study,
• Do students learn braille contractions one author was available for assistance
more efficiently with Braille Tutor? with the assessments and procedures of
• Does Braille Tutor provide explana- the study, and a different author was
tions that are clear? available for technical assistance.
• Is Braille Tutor fun to use?
STUDYDESIGN
Method
The study was approved by the Institu-
PARTICIPANTS tional Review Board of the Wexford
Thisstudyinvolved10studentswhowere Institute. It employed a single-subject,
instructed by 7 teachers of students with adapted alternating treatments design
visual impairments. Inclusion criteria for with a probe at the end of each phase.
student participants were that the student: Experimental control was demonstrated
(a)wasabraillereaderwhodidnotknow primarily within participants but was also
311
©2016AFB,AllRightsReserved JournalofVisualImpairment&Blindness,September-October2016
CE Article
demonstrated across participants. Teach- During a session with the device, the
ers were also interviewed to determine promptonthescreenwaspresentedtothe
attitudes about the device. student via the screen reader. For this
All participants began at the baseline study, the prompt was a single word, al-
phase. After baseline, participants were though the device can also present sen-
assigned to one of two groups: TVI Only tences. After hearing the word spoken by
or TVI(cid:3)Tutor. Participants in the TVI thescreenreader(theprompt),thestudent
Only group taught a list of words to their brailledthewordusingabraillekeyboard
student in their typical manner. Partici- or a QWERTY keyboard that permitted
pants in the TVI(cid:3)Tutor group were in- six-key entry. The Braille Tutor website
structed to teach the students in the man- contains different units that focus on dif-
ner they would typically teach and to ferent types of contractions (for instance,
reinforce instruction by allowing the stu- single-letter, whole-word contractions or
dent to use Braille Tutor as often as they short-form words). Some units focus on
desired,oraminimumoftwosessionsper the contractions in isolation; some focus
week. on the contractions in the context of sen-
After the teachers finished their first tences.Thisstudyonlyincludedunitsthat
assigned phase with students (TVI Only focused on contractions in isolation. For
or TVI(cid:3)Tutor), they completed a probe example, a computer programmer set up
using the adapted Assessment of Braille an individualized unit for each student
LiteracySkills(ABLS,Koenig&Farren- participant that was based on the iden-
kopf, 1995) tool and then switched to the tifiedlistofcontractedwordsinorderto
other phase. For example, if the teacher specifically control which contractions
was originally assigned to the TVI Only were introduced during each phase of
phase,atthecompletionofthatphase,the the study. Typically, however, a teacher
student would be probed and would then would choose a unit from a prepro-
start the TVI(cid:3)Tutor phase. The student grammed list that most closely aligned
would be probed again to assess mainte- with the types of contractions the stu-
nance of the contractions learned in the dent was currently learning. What sets
first interventional phase. BrailleTutorapartfromotherdevicesis
that the program analyzes the type of
INSTRUMENTSANDTOOLS brailling mistake made by a student and
BrailleTutor offers specific feedback based on the
Braille Tutor is an Internet-based tool particular mistake rather than providing
that uses a form of artificial intelligence a stock response. For instance, if a stu-
known as adaptive computer instruction. dent was given the prompt “not” and
Eachpersonwhousesithasauniqueuser brailledeachletter—n,o,t—BrailleTu-
name and password. The user logs into tor reminded the student that there was
the Braille Tutor website using a com- a single-cell whole-word contraction
puter with a screen reader, a refreshable that could have been used. This type of
brailledisplay,andabraillekeyboardora feedback can be compared to a student
QWERTY keyboard that permits six-key who was given the same prompt but
entry. brailled the contraction upside down
312
JournalofVisualImpairment&Blindness,September-October2016 ©2016AFB,AllRightsReserved
CE Article
Figure1. TheBrailleTutorusesatypeofartificialintelligencereferredtoasadaptivecomputer
instruction.
(dots 1, 3, 5, 6). In the case of upside- words (such as br (cid:2) braille). Other
down braille, the device would tell the contractions were not assessed, since
student that the dots had been flipped the prototype of the Braille Tutor that
top to bottom (see Figure 1). was used for this project only provided
practice with these two types of con-
Assessmentofbrailleliteracyskills tractions. These words were brailled
An adapted version of the ABLS tool, into a word list. Teachers were asked to
which was also used in the Alphabetic administer the list without specific
Braille Contracted (ABC) Braille Study prompts (for example, “Sound it out, b,
(Wall Emerson et al., 2009), was used to b, b”), although general prompts of en-
establish prior student knowledge of couragement were permitted (such as
contractions and baseline). Although “Just do your best”) to reduce student
the format of the tool was the same frustration. Teachers were also in-
as the tool used in the ABC Braille structed to not give any feedback on
Study,thechecklistweusedinthepres- student performance (for instance,
ent study only contained the one-cell, “You got it right”). The ABLS was ad-
whole-word contractions that are repre- ministered as a pretest to establish a list
sented by a single letter of the alphabet of target words for each participant;
(for example, b (cid:2) but) and short-form theselistsalsoservedasthefirstdatapoint
313
©2016AFB,AllRightsReserved JournalofVisualImpairment&Blindness,September-October2016
CE Article
Table1
Studentdemographicdata.
Numberof
Educational target
Name Gender Grade Age Ethnicity Eyecondition(s) placement words
Adam M Pre-K 4:11 Caucasian Opticnervedamage Itinerant 14
Bobby M K 5:10 Hispanic Leber’scongenitalamaurosis Residential 8
Curt M K 5:4 Mixedrace Septo-opticdysplasia Residential 8
Ellen F K 6:8 AfricanAmerican Alagillesyndrome Residential 8
Gina F 2 7:5 Caucasian Coloboma Itinerant 22
Hal M 2 7:9 Caucasian Opticnervedamage Itinerant 8
Lisa F 5 10:2 Hispanic Leber’scongenitalamaurosis Residential 20
Mark M 4 10:7 Hispanic Retinitispigmentosa Itinerant 20
Rachel* F 8 13:9 Trauma Itinerant 20
Sue F 9 14:11 Caucasian Rodconedystrophy Itinerant 30
K(cid:2)kindergarten.
*Racheldidnotdiscloseherethnicity.
inbaseline.Thetoolwasalsoadministered simply because they received more envi-
after each phase as a probe. ronmental exposure to them. Words were
organized by frequency using the Stan-
Wordlists dard Frequency Index (SFI; Breland,
After administering the adapted ABLS 1996), which is based on word frequency
once, the teachers were asked to select a asmeasuredbytheEducator’sWordFre-
list of the contracted words each student quency Guide (Zeno, Ivens, Millard, &
missed in the ABLS. The words the Duvvuri, 1995). The first author rank-
teacher selected were the target words ordered the words according to their SFI
that were used for the remainder of the and assigned words with an odd rank or-
study.Oncetheteacherselectedthesetar- der number to the TVI Only group and
get words, the first author confirmed that words with an even rank order number to
thewordsselectedbytheteacherswerein the TVI(cid:3)Tutor group.
fact words that the student had missed.
The authors recommended that teachers PHASES
choose 20 target words, but the teachers Baseline
could choose a different number if they In baseline, the adapted ABLS was ad-
wished. They were instructed to “choose ministered a minimum of two more times
enough words so your student will not after target word selection. This testing
learntheentirelistinadayortwo,butnot was done to establish that the students
so many that the student will be over- definitivelydidnotknowanyofthetarget
whelmed.” The number of target words words.Baselinewasconducteduntilthere
for each student varied based on age and were at least three data points with zero
ability (see Table 1). celeration.
The words were divided into two
groups and were matched based on their Intervention:TVIOnly
word frequency to ensure that students Half of the student participants were as-
did not learn some words more quickly signed to the TVI Only phase as the first
314
JournalofVisualImpairment&Blindness,September-October2016 ©2016AFB,AllRightsReserved
CE Article
interventionalphaseafterbaseline.Inthis the participant began using it at the end
phase,theteacherwasinstructedtobegin of the week or had met the criteria after
teaching only the target words to the TVI one session during a particular week.
Onlygroup.Theteacherwasinstructedto The average session with the Braille
teach the words in the usual manner and Tutor lasted approximately 15 minutes,
at the typical pace, and was asked to test but varied based on the number of
the words being used in this phase by words on each student’s individualized
having the student read the words using list. This phase was considered com-
ABLS after each instructional period. plete when the student identified the
Teachers were instructed to send the re- target words with 100% mastery in
sults of testing to the first author on the three consecutive sessions.
same day they completed the assessment.
This phase was considered complete Probes
when the student identified the target
Between the two intervention phases
words with 100% mastery in three con-
and after the completion of the second
secutive sessions.
intervention phase, teachers were in-
structed to administer the adapted
Intervention:TVI(cid:2)Tutor
ABLS so the authors could see the stu-
Half of the student participants were
dents’ total contraction knowledge at
assigned to the TVI(cid:3)Tutor phase as the
different points throughout the study.
firstinterventionalphaseafterbaseline.In
This ongoing monitoring helped control
this phase, the teacher was instructed to
for maturation as a threat to internal
begin teaching only the target words to
validity and allowed the authors to as-
the TVI(cid:3)Tutor group. The words were
sess maintenance of words learned in
taught in the typical manner and pace.
the first interventional phase.
The teachers were asked to test the stu-
dent on the selected target words on days
they worked with the student or on days INTER-RATERRELIABILITY
the student used Braille Tutor. If the stu- The teachers were the primary coders of
dent used the device on a day when the studentprogress.Toensureteacherswere
teacher was not present, the teacher was following instructions related to student
askedtohaveaclassroomteacherorpara- promptingandfeedbackaswellascoding
educator who had received training in student errors correctly, teachers were in-
conducting the assessment evaluate the structed to record a video of a session
student on the target words. Participants with a student at least once during base-
were required to use the device a mini- line and every fifth time during interven-
mum of two times per week, but were tion. The video recordings were likewise
permitted to use it as often as they liked. coded by the first author. In baseline,
On average, participants used the de- inter-rater reliability was collected on
vice three times per week, with a range more than 33% of sessions. In all cases,
of one to five times. In all but two teachers were 100% reliable in coding
instances, when the device was only student responses on the ABLS. Since
used one time per week, it was because they were so reliable in baseline, in order
315
©2016AFB,AllRightsReserved JournalofVisualImpairment&Blindness,September-October2016
CE Article
to reduce the amount of time spent in attended a residential school and the re-
study activities versus instruction, the mainder attended their local schools. The
percentage of sessions monitored was students ranged from pre-kindergarten to
reduced to 23%. All reliability checks ninth grade, and they had various eye
held at 100%. conditions.
TEACHERINTERVIEWS BASELINE,INTERVENTION,ANDPROBES
Dataweregraphedandanalyzedvisually.
Following the student’s work with the
In Figures 2 and 3, instructional days run
Braille Tutor, the teachers were inter-
along the x-axis. The y-axis indicates the
viewedbythesecondauthor.Interviews
percentage of contracted words the stu-
lasted 20 to 30 minutes and were de-
dents learned. Connected data points rep-
signed to understand the experience the
resent data that was collected on consec-
teacherandstudenthadusingthedevice
utive days. Data points that are not
and the teacher’s thoughts on how it
connected indicate a break between in-
could be expanded and improved. A
structional days. The dotted vertical lines
structured interview containing 14
indicate a phase change. In baseline, the
questions was used.
student participants were assessed on
contraction knowledge using the adapted
Results
ABLS. The percentage of the total target
Data collection occurred from October
words(wordsinboththeTVIOnlyphase
2013 to March 2014. Ten students com-
and words in the TVI(cid:3)Tutor phase) the
pleted all phases of the study.
studentreadcorrectlywasgraphed.Inthe
first intervention phase, the student was
DEMOGRAPHICDATA
assessed on the target words for only that
Teachers
phase (TVI Only or TVI(cid:3)Tutor). The
Seven teachers completed the study. All
percentage the student read correctly was
were Caucasian: one (14.3%) was male
graphed. The probe following the first
and six (85.7%) were females. All but
phase allowed the authors a measure of
one self-reported extreme comfort with
maintenance and generalization.
a computer. Teaching experience
rangedfrom1to16yearswithamedian TVIOnlyphasefirst
of11.5years.Five(71.4%)oftheteach-
All of the students who were in the TVI
ers were itinerant and 2 (28.6%) were
Only phase first made progress. Three
employed at a residential school. Five
students who were in the same classroom
of the teachers had 1 student who par-
with the same classroom teacher (who
ticipated in the study, while 1 had 2
was a teacher of visually impaired stu-
students in the study, and 1 had 3 stu-
dents), made slow and steady progress,
dents in the study.
acquiringonenewtargetwordeachweek
(Curt, Bobby, and Ellen). This contrasts
Students with Sue, who made quick progress after
Table 1 reports the demographic data for an initial period of learning. Lisa learned
thestudents.Four(40.0%)ofthestudents very quickly, but did not retain all the
316
JournalofVisualImpairment&Blindness,September-October2016 ©2016AFB,AllRightsReserved
CE Article
Figure 2. Participants who completed the sequence: Baseline, TVI Only, Probe, TVI(cid:3)Tutor,
posttest.
words she had learned until she had prac- Most students made a quick jump after
ticed with her teacher several times (see the first day of instruction reinforced by
Figure 2). the Braille Tutor. One student jumped to
20% correct (Rachel) but did not have a
TVI(cid:2)TutorPhaseFirst steep increase until the fifth session. It
Students who began the study in the should be noted that the fourth session
TVI(cid:3)Tutor phase also made progress. waswhenRachelactuallyusedthedevice
317
©2016AFB,AllRightsReserved JournalofVisualImpairment&Blindness,September-October2016
CE Article
for the first time, so the steep jump oc- their students was the opportunity to use
curred right after its introduction. technology, breaking the monotony of
Gina’s teacher ended the TVI(cid:3)Tutor workingwiththeirteachers.Afewteach-
phase just before the student’s holiday ers reported their students did not enjoy
break. She probed the student when the using Braille Tutor after the initial nov-
student returned after three weeks. When eltyworeoff.Reasonscitedincludedthat
the evaluator realized that a three-week the device did not vary the order of pre-
break had occurred, she asked that Gina sentation, that the feedback was limited,
return to the TVI(cid:3)Tutor phase for two andthattherewassignificanttimeneeded
more data points at 100% before switching between pressing “Check My Work” and
totheTVIOnlyphaseinordertopreventa being provided with the next question.
phase change from occurring after a gap in Thesereasons,thoughvalid,werenotun-
data collection (see Figure 3). der the control of the authors, since the
Braille Tutor prototype was limited in its
Pretestandposttest abilities and many schools had slow In-
All students in the study made prog- ternet connections that led to time lags.
ress from pretest to posttest on the Typically, the device requires less than
Adapted ABLS, which measured one second to process the entered infor-
student contraction knowledge using a mation and respond.
word list. On average, students learned Almost all students learned to navigate
16 (range 9 to 30) braille contractions Braille Tutor independently once the
during the study. teacher completed any “setup” (such as
setting up the refreshable braille display).
TEACHERINTERVIEWS Teachers of younger students felt the lan-
The second author conducted telephone guage used in the feedback was more
interviews with the teachers following advanced than the child’s level. Several
the students’ work with Braille Tutor. teachers, especially those of older stu-
The teachers reported that after initial dents, reported that the students quickly
instruction with the device’s functions, becameboredwiththefeedback,whichdid
the majority of students were able to use not vary. They suggested, for example, us-
it independently with minimal assistance ingavarietyofreinforcersoundsorphrases
from the teacher. The older the student, such as “Way to go!” or “You aced this
the more likely that the student used the one.” Several of the students in the study
device with greater independence. The were dual-media users, and their primary
majority of teachers felt students would literacy medium was print, not braille.
be able to use Braille Tutor in the general Teachers reported that these students navi-
education classroom with minimal sup- gated Braille Tutor visually, not auditori-
port. The majority of students were pos- ally. The amount of information on the
itive about using the device, with one screenwasproblematicforonevisualuser.
second-grader even writing the Braille The teacher suggested that a screen that
Tutor a letter to tell “him” how much she containedonlythebrailleconfigurationand
enjoyedworkingwithhim.Severalteach- theneededbuttonsinalargerfontwouldbe
ers reported that the motivating factor for helpful for dual-media users.
318
JournalofVisualImpairment&Blindness,September-October2016 ©2016AFB,AllRightsReserved