CASE STUDY
Alameda Police Department
36% Reduction in Use of Force after
Implementation of Training and Body-Worn
Camera Analytics
Tejas Shastry, Ph.D.*
BACKGROUND
Recent research shows that the use of body-worn cameras (BWC) alone does not signiꢀcantly improve police performance and accountability. Speciꢀcally, there remains substantial uncertainty about whether BWC can alone reduce oꢁcer use of force. Cynthia Lum, et al. have shown a lack of conclusive evidence that police behaviors change solely from the use of body worn cameras and instead suggest that the use of BWC for coaching, training, and accountability could successfully deliver on these promises.1 AUTHOR CONCLUSION:
These results provide strong evidence that the implementation of strategic communication training and automated BWC analytics can reduce the number of use of force occurrences, improve oꢁcer professionalism and, as a result, improve civilian compliance. Moreover, we ꢀnd that when oꢁcers use unprofessional language in an eꢂort to de-escalate, this language does not reduce civilian noncompliance.
PLAIN LANGUAGE SUMMARY:
Automated BWC analytics provides a method for departments to review 100% of their BWC videos to surface positive interactions that reinforce professionalism training and correct risky behavior as it occurs, fulꢀlling the promise of BWC. In this study, we demonstrate that the use of BWC analytics combined with professionalism training leads to substantial reduction in use of force occurrences (across de minimis through lethal use of force) as well as an increase in positive behavioral changes in both oꢁcers and the community members they serve.
This study presents an analysis of BWC data from the
police department of Alameda, California comparing the
ꢀrst half (H1) of 2021 vs ꢀrst half of 2022. In H2 of 2021, Alameda PD Chief Nishant Joshi implemented several
training programs centered around strategic
communication and began use of Truleo’s automated BWC
analysis software. Prior to Truleo, Alameda PD was
leveraging Sergeants to conduct random audits that
resulted in reviewing less than 1% of videos. After training and automated BWC analytics were implemented, we
observed the following changes in the Alameda PD data:
• 36% decrease in use of force
• 30% decrease in unprofessional oꢀcer language
• 17% increase in oꢀcer explanations
• 12% decrease in civilian non-compliance
• 99% of interactions were at a standard or high
professionalism level
I commend our leadership team,
from command staꢀ to front-line
supervisors to the oꢁcers on the
front line. They all bought into the
mission and leveraged BWC analytics
as a coaching tool to help us create
better outcomes with our community.
The data helped us celebrate the
99% of interactions that were
professional while learning how to
improve from the other 1%.
Chief Nishant Joshi
ALAMEDA POLICE DEPARTMENT
01CASE STUDYALAMEDA PD
EXHIBIT 1
C
A
SE S
TUD
Y
Alameda P
olice Department
Implementation of T
raining
and B
W
C Analytics
T
ABLE 1.
Summary of metrics from B
W
C analysis demonstrating improvements in professionalism.
Instances of
Oꢀcer Explanation
2021 H1
1.4 2022 H1
1.6 +14.3%
50 .6
59 .2
+17%
0 .1
7 0
.1
5 -11.8%
1.25 0
.88
-29
.6%
4.27 5.42
+26.9%
0 .69
0 .44
-36.2%
P
ercentage
Change
W
ord Count of
Oꢀcer Explanation
Instances of
Civilian Noncompliance
Instances of
Oꢀcer Rudeness
Instances of
Civilain Rudeness
Instances of
Use of Force
A
s man
y cities struggle with incr
easing t
ensions
bet
w
een polic
e and the c
ommunities they ser
v
e
,
r
esear
ch has suggest
ed tha
t mor
e pr
of
essional
beha
vior on the par
t of polic
e oꢁc
ers can lead t
o
mor
e eꢂ
ec
tiv
e policing
. I
n par
ticular
, r
esear
ch
suggests tha
t oꢁc
ers using polit
e language and
pr
o
viding e
xplana
tion of ac
tions bef
or
e they
oc
cur
can lead t
o
mor
e positiv
e
in
t
er
ac
tions
with
civilians
.
2,3
I
ncr
eases in positiv
e in
t
er
ac
tions
bet
w
een polic
e and c
ommunit
y members ha
v
e
also been sho
wn t
o lead t
o a gener
al impr
o
v
e
-
men
t in civilian c
omplianc
e with polic
e
.
4 F
r
om
the star
t of Chief
Joshi
’
s t
enur
e a
t
A
lameda
PD in
June of 202
1 , sev
er
al
polic
y changes w
er
e
implemen
t
ed:
1 .
L
eadership acr
oss the depar
tmen
t (c
ommand
staꢂ
thr
ough
fr
on
t
-line
super
visors)
pr
ior
-
itiz
ed body
-
w
or
n camer
a ac
tiv
a
tions f
or ev
er
y
public enc
oun
t
er and c
odiꢀed polic
y c
onse
-
quenc
es f
or high failur
e of ac
tiv
a
tion.
3. L
eadership implemen
t
ed an o
v
er
1 0x incr
ease
in r
equir
ed tr
aining hours
, fr
om 30 hours
ev
er
y
2 y
ears t
o
330
hours
ev
er
y
2 y
ears
.
A
c
or
e emphasis of the additional tr
aining w
as
de
-
escala
tion and c
ommunica
tion tr
aining
.
T
his lar
ge change pr
o
vided an oppor
tunit
y
f
or new tr
aining policies t
o ha
v
e a sig
niꢀcan
t
inꢃuenc
e on new r
ecruits
.
4 .
I
n Januar
y of 2022, A
lameda implemen
t
ed
T
ruleo
B
WC analytics
.
H
ighly
pr
of
essional
in
t
er
ac
tions w
er
e sur
fac
ed
b
y
T
ruleo
’
s sof
t
w
ar
e
,
and
c
ommand staꢂ
sen
t depar
tmen
t
-
wide
c
ommunica
tions c
ong
r
a
tula
ting good oꢁc
er
beha
vior as a w
a
y t
o r
einf
or
c
e str
a
t
eg
ic
c
ommunica
tion tr
aining
.
5. F
ollo
wing the implemen
ta
tion of B
WC
analytics
, S
er
gean
ts a
t A
lameda PD began
using a pr
ecision-
guided audit w
or
kꢃo
w (in
c
on
tr
ast t
o r
andom audits).
T
ruleo
’
s B
WC
analytics w
ould aut
oma
tically sur
fac
e
in
t
er
ac
tions with pot
en
tial r
isk
y beha
vior
,
allo
wing the r
eview
er
t
o go t
o the pr
ecise
momen
t of oꢁc
er unpr
of
essional language
and
det
er
mine
if
it
w
as
r
isk
or
not
,
enabling
the r
eview of key momen
ts acr
oss hundr
eds
of videos in the time a single video w
ould take
t
o r
eview
.
2. A
s of D
ec
ember 2022, 67% of A
lameda pa
tr
ol
oꢁc
ers ar
e new
, a net change of 2
1 pa
tr
ol
oꢁc
ers in addition t
o 8 of
14 pr
omot
ed
S
er
gean
ts and
4 of 6 pr
omot
ed Lieut
enan
ts
.
Each v
alue is nor
maliz
ed t
o audio hours of in
t
er
ac
tion, wher
e an in
t
er
ac
tion audio hour c
on
tains speech
fr
om either an oꢁc
er or a civilian. A
n instanc
e of a metr
ic is a single sta
t
emen
t b
y an oꢁc
er or civilian tha
t
indica
t
es tha
t label
.
02 C
A
SE S
TUD
Y
ALAMED
A PD
C
A
SE S
TUD
Y
Alameda P
olice Department
Impact of P
olicy Measured
by B
W
C Analytics
Impact of P
olicy Measured by
B
W
C Analytics
Chief Joshi
began his
t
enur
e a
t
A
lameda
P
olic
e
D
epar
tmen
t in June of 202
1 .
T
o make a fair
c
ompar
ison,
this
study
c
ompar
es
da
ta
fr
om
Januar
y
1 st - June 30th of 202
1 (H
1 202
1 ) t
o the
same time per
iod c
ompr
ising H
1 2022. I
n this
w
a
y
,
an
y
seasonal
changes
in
oꢁc
er or
civilian
beha
vior
,
as w
ell
as
sig
niꢀcan
t
w
or
ld ev
en
ts
and
their eꢂ
ec
t
, ar
e disr
egar
ded
.
T
his study c
onduc
t
ed B
WC analy
sis using
machine
-aut
oma
t
ed speech
r
ec
og
nition, speaker
diar
iza
tion, oꢁc
er v
oic
e and language
iden
tiꢀca
tion, and na
tur
al language pr
oc
essing
.
A
ll machine
-lear
ned models in this pr
oc
ess used
der
iv
a
tions of tr
ansf
or
mer
-based deep
-lear
ning
models
, which understand language based on
local c
on
t
e
x
t
.
5 T
able
1 sho
w
s a summar
y of key metr
ics
measur
ed fr
om B
WC analy
sis dur
ing the t
w
o
6-mon
th time per
iods and the o
v
er
all per
c
en
tage
change bet
w
een the time per
iods
. B
elo
w w
e
summar
iz
e the key r
esults
.
C
ompar
ed t
o H
1 202
1 , H
1 2022 sho
w
ed a 55%
incr
ease in
r
a
w B
WC f
ootage dur
a
tion and a
1 0
4 %
incr
ease in B
WC in
t
er
ac
tion dur
a
tion. B
WC f
oot
-
age dur
a
tion incr
ease suggests tha
t oꢁc
ers ar
e
tur
ning
on
their
camer
as
mor
e
of
t
en
than
pr
evi
-
ously
. B
WC in
t
er
ac
tion dur
a
tion incr
ease indi
-
ca
t
es tha
t the times wher
e oꢁc
ers ar
e tur
ning on
their camer
as ar
e o
v
er
whelmingly ones tha
t ha
v
e
oꢁc
er and civilian speech r
a
ther than just silenc
e
or backg
r
ound noise
.
T
he depar
tmen
t polic
y emphasis on main
taining
high numbers of camer
a ac
tiv
a
tions likely r
esult
-
ed in the obser
v
ed incr
ease in B
WC in
t
er
ac
tions
.
Comparative Improvement in
Professionalism
B
et
w
een H
1 202
1 and H
1 2022, this study
obser
v
ed a
1 7%
incr
ease
in
the
number
of w
or
ds
of e
xplana
tion oꢁc
ers w
er
e deliv
er
ing per hour
of B
WC f
ootage on a
v
er
age
. I
n c
onc
er
t with this
incr
ease in e
xplana
tion, w
e obser
v
ed a
1 2%
r
educ
tion in civilian non-
c
omplianc
e bet
w
een H
1 202
1 and H
1 2022.
W
e pr
opose a deꢀnition f
or a
“
standar
d
pr
of
essional
”
in
t
er
ac
tion as one with no
unpr
of
essional
oꢁc
er
language
,
and a
“
highly
pr
of
essional
”
in
t
er
ac
tion as ha
ving thr
ee
c
onditions: (
1 ) no unpr
of
essional language fr
om
the oꢁc
er
, (2) 25 w
or
ds or
mor
e of e
xplana
tion
used bef
or
e a use of f
or
c
e ev
en
t
, and (3) using a
t
least one instanc
e of polit
e language
. U
sing this
deꢀnition, the r
esulting per
c
en
tage of highly
pr
of
essional in
t
er
ac
tions incr
eased fr
om
1 0% in
H
1 202
1 t
o
1 2% in H
1 2022, a 20% r
ela
tiv
e
incr
ease in the number of highly pr
of
essional
in
t
er
ac
tions
. O
v
er both per
iods
, the number of
pr
of
essional in
t
er
ac
tions (standar
d pr
of
essional +
highly pr
of
essional) w
as
98% and 99%,
r
espec
tiv
ely
, indica
ting a 50% r
educ
tion in
unpr
of
essional in
t
er
ac
tions
.
T
his study also obser
v
ed a 36% decr
ease in
language indica
tiv
e use of f
or
c
e fr
om H
1 202
1 t
o
H
1 2022, which includes all t
ypes of use of f
or
c
e
.
A
lameda PD r
epor
ts all use of f
or
c
e and c
onduc
ts
f
ollo
w
-up in
v
estiga
tions in inciden
ts with injur
ies
or c
omplain
ts
. I
n these f
ollo
w
-up in
v
estiga
tions
,
the agenc
y obser
v
ed a 70% r
educ
tion o
v
er the
same time per
iod
, c
or
r
obor
a
ting the tr
end
obser
v
ed via B
WC analytics
.
T
he measur
e of use of f
or
c
e thr
ough language
can c
ome fr
om
either the civilian or the
oꢁc
er as
an indica
tion a use of f
or
c
e is oc
cur
r
ing
.
I
n par
allel
, this study obser
v
ed a 30% decr
ease in
oꢁc
er use of unpr
of
essional language bet
w
een
H
1 202
1 and H
1 2022, despit
e a 27% incr
ease in
civilian usage of unpr
of
essional language
c
ompar
ing the same time per
iods
. Her
e
,
unpr
of
essional language is deꢀned as the use of
pr
ofanit
y
, insults
, and thr
ea
ts
.
03 C
A
SE S
TUD
Y
ALAMED
A PD
C
A
SE S
TUD
Y
Alameda P
olice Department
Unprofessional language fails
to de-escalate scenarios
I
n addition t
o the c
ompar
a
tiv
e br
oad r
esults of
incr
eased pr
of
essionalism in the A
lameda P
olic
e
D
epar
tmen
t
, this study also analyz
ed in
t
er
ac
tions
acr
oss the same time per
iod t
o understand oꢁc
er
use of unpr
of
essional language
.
O
ꢁc
ers t
end t
o deplo
y unpr
of
essional language
as an a
tt
empt t
o
de
-
escala
t
e a nonc
omplian
t
civilian. Ho
w
ev
er
, in analyzing these in
t
er
ac
tions
,
w
e ꢀnd tha
t the use of unpr
of
essional language
fails t
o decr
ease civilian nonc
omplianc
e
.
T
he per
c
en
tage of civilian language in r
ec
or
ded
B
WC c
on
taining instanc
es of nonc
omplian
t
language (the nonc
omplianc
e r
a
t
e) acr
oss all
in
t
er
ac
tions w
as 2.5%.
A
s F
igur
e
1 sho
w
s
, the r
a
t
e of civilian nonc
ompli
-
anc
e incr
eases af
t
er ev
en one oc
cur
r
enc
e of
unpr
of
essional oꢁc
er language
.
Data Collection and Analysis
T
he da
ta f
or this study w
er
e c
ollec
t
ed via an API
pr
o
vided b
y the depar
tmen
t
’
s evidenc
e manage
-
men
t sy
st
em.
V
ideos deliv
er
ed via the API w
er
e
f
ed thr
ough an audio e
x
tr
ac
tion pipeline such
tha
t only the audio inf
or
ma
tion of the video w
as
r
etained in t
empor
ar
y memor
y f
or analy
sis
. A
udio
w
as f
ed thr
ough machine
-lear
ned speech r
ec
og
-
nition models t
o obtain w
or
ds and timestamps of
v
alid speech in the audio ꢀle
.
V
ideo ꢀles tha
t
c
on
tained only noise or silenc
e
w
er
e discar
ded
dur
ing the audio analy
sis pr
oc
ess
.
Data Collection and Analysis
M
an
y ac
cur
ac
y metr
ics ar
e a
v
ailable f
or B
WC anal
-
y
sis
models
, including
pr
ecision,
r
ecall
,
f-beta,
and c
on
v
en
tional
“
ac
cur
ac
y
”
. F
or
this study
,
pr
eci
-
sion w
as chosen as the metr
ic of ac
cur
ac
y t
o be
optimiz
ed acr
oss all models
.
B
y na
tur
e
,
B
WC videos will
not captur
e
all ev
en
ts
as the
quan
tit
y of video
depends on an oꢁc
er
ph
y
sically tur
ning on their B
WC.
T
his
puts an
undet
er
mined limit on r
ecall
, which is the abilit
y
of a model t
o c
or
r
ec
tly iden
tify all ev
en
ts or labels
tha
t ma
y ha
v
e oc
cur
r
ed in all k
no
wn in
t
er
ac
tions
.
La
t
e ac
tiv
a
tion
and
ear
ly t
er
mina
tion of
B
WC
video b
y the oꢁc
er can e
xac
er
ba
t
e these issues
.
W
hile a 3x incr
ease of nonc
omplianc
e oc
curs
af
t
er ev
en 2 oc
cur
r
enc
es of unpr
of
essional
language
,
4 or mor
e oc
cur
r
enc
es incr
ease non
-
c
omplianc
e b
y 8x t
o 20x t
o a maximum of 35%
nonc
omplianc
e
.
T
his means f
or in
t
er
ac
tions with
6 or mor
e instanc
es
of unpr
of
essional language
fr
om an oꢁc
er
,
1 out of 3 civilians is likely t
o be
nonc
omplian
t
.
W
e not
e tha
t it cannot be c
oncluded fr
om this
da
ta tha
t the oꢁc
er
’
s use of unpr
of
essional
language w
as the cause of fur
ther nonc
ompli
-
anc
e
, but these da
ta clear
ly demonstr
a
t
es tha
t
unpr
of
essional language does not assist an
oꢁc
er in obtaining bett
er civilian c
omplianc
e
.
Methods
O
v
er
all
,
the
incr
ease
in
oꢁc
er
e
xplana
tion
and
r
educ
tion in oꢁc
er unpr
of
essional language
suggests a c
onc
er
t
ed eꢂ
or
t fr
om oꢁc
ers t
o
impr
o
v
e the pr
of
essionalism of civilian in
t
er
ac
-
tions with str
a
t
eg
ic language
, r
ew
ar
ded with a
decr
ease in civilian nonc
omplianc
e and
decr
eased need f
or use of f
or
c
e fr
om oꢁc
ers
.
04 C
A
SE S
TUD
Y
ALAMED
A PD
C
A
SE S
TUD
Y
Alameda P
olice Department
Oꢀcer Identification and Accuracy
Subsequen
t analy
sis on the audio w
as c
onduc
t
ed
t
o diar
iz
e the audio in
t
o anon
ymous speakers
,
and a language
-based model w
as used t
o iden
tify
which of the anon
ymous speakers w
as likely t
o be
the oꢁc
er w
ear
ing the camer
a. I
n t
esting
, this
model per
f
or
med with an 85% lev
el of pr
ecision.
T
he r
emaining speakers w
er
e agg
r
ega
t
ed in
t
o
civilian audio
.
T
he same model w
as used acr
oss all
analy
ses
, enabling c
onꢀden
t c
ompar
isons of
r
ela
tiv
e diꢂ
er
enc
es ack
no
wledg
ing tha
t absolut
e
numbers ma
y be aꢂ
ec
t
ed b
y pr
ecision er
r
ors
.
Natural L
anguage Processing Events,
L
abels, and Accuracy
A
f
t
er oꢁc
er / civilian iden
tiꢀca
tion, the r
esulting
t
e
x
t
w
as
analyz
ed
via na
tur
al language
pr
oc
essing (NLP) models tha
t iden
tiꢀed c
er
tain
ev
en
ts and labels
.
NLP models w
er
e tr
ained fr
om da
ta sets wher
e
humans tagged seg
men
ts of tr
anscr
ibed t
e
x
t
with r
elev
an
t ev
en
ts and labels
, with a tar
get
in
t
er
-annota
t
or ag
r
eemen
t lev
el e
x
c
eeding 90%.
E
v
en
ts w
er
e output via an in
t
en
t classiꢀca
tion
model tak
ing a seg
men
t of t
e
x
t fr
om
1 t
o 50
w
or
ds fr
om a single utt
er
anc
e of a single speaker
and outputting a single label
. Labels w
er
e output
via either an in
t
en
t classiꢀca
tion model or a
named
en
tit
y
r
ec
og
nition model
tha
t
tags
one or
mor
e w
or
ds with a speciꢀc en
tit
y tag
.
E
v
en
ts w
er
e deꢀned as language tha
t indica
t
ed an
ev
en
t
oc
cur
r
ed
in
a
B
WC
in
t
er
ac
tion,
whether
fr
om
the civilian or the oꢁc
er
.
T
he t
w
o ev
en
ts r
elev
an
t t
o
this study w
er
e
“
use of f
or
c
e
”
and
“
nonc
omplianc
e
”
.
Language f
or both of these ev
en
ts c
ould oc
cur via
either an oꢁc
er or civilian. U
se of f
or
c
e language
w
as c
ollec
t
ed t
o
c
o
v
er
the en
tir
e r
ange
of use
of
f
or
c
e
, fr
om de minimis f
or
c
e thr
ough lethal f
or
c
e
.
S
ome
e
xamples of
language indica
ting
use of
f
or
c
e
include
“
taser deplo
y
ed
”
on behalf of the oꢁc
er or
“
don
’
t tase me
”
on behalf of the civilian. Similar
ly
,
“
y
ou
’
r
e
still
r
efusing
t
o c
ooper
a
t
e
”
and
“I
w
on
’
t
c
omply
”
r
epr
esen
t the language of
“
nonc
omplianc
e
”
fr
om the angle of the oꢁc
er and
civilian, r
espec
tiv
ely
.
Labels w
er
e classiꢀed as pr
of
essional or rude
language
. P
r
of
essional language
w
as
dissec
t
ed
in
t
o the use of
“
polit
e
”
language and the use of
e
xplana
tion. P
olit
e language includes the use of
f
or
malit
y
, g
r
a
titude
, and polit
eness
, such as
“
M
r
. /
Sir / M
a
’
am
”
,
“
T
hank
Y
ou
”
, and
“P
lease
”
, r
espec
tiv
ely
.
Explana
tion is deꢀned as instanc
es of oꢁc
ers
e
xplaining
t
o
a
civilian
wh
y they
ar
e
tak
ing
a
c
er
tain
ac
tion
(such as a
fr
isk
,
det
en
tion,
or ar
r
est)
using mor
e than 25 w
or
ds
.
Rude language is deꢀned as pr
ofanit
y (whether
dir
ec
t
ed or not
-
“f***
y
ou
”
or
“
s***”
, r
espec
tiv
ely),
insults (r
acial slurs), or thr
ea
ts (“I
’
m going t
o k
ill
y
ou
”).
I
n t
esting
, the NLP pr
ecision f
or this study w
as o
v
er
90% f
or
“
use of f
or
c
e
”
,
“
nonc
omplianc
e
”
,
“
thr
ea
t
”
,
“
pr
ofanit
y
”
, “
polit
eness
”
,
and “
e
xplana
tion
”
.
A
dditionally
, ev
en in a r
ec
or
ded in
t
er
ac
tion, it is
possible tha
t an ev
en
t oc
curs without v
er
bal cues
,
such as a use of f
or
c
e
.
K
no
wing tha
t all in
t
er
ac
tions ma
y not be
captur
ed
, a c
ompar
a
tiv
e analy
sis can still be
c
onduc
t
ed assuming the char
ac
t
er
istics list
ed
abo
v
e do not sig
niꢀcan
tly
change
. I
n these
c
ompar
a
tiv
e
analy
ses
,
pr
ecision
(the
number
of
true positiv
es disc
oun
ting false positiv
es
,
ig
nor
ing false nega
tiv
es) is the best metr
ic as
r
ela
tiv
e v
alues can be c
ompar
ed with c
onꢀdenc
e
.
05 C
A
SE S
TUD
Y
ALAMED
A PD
C
A
SE S
TUD
Y
Alameda P
olice Department
A
uthor A
ꢀliations
Citations
About T
ruleo
*
T
ruleo Inc.
1 E Erie S
treet, Suite 525-2246
Chicago
, IL 60611
Risk and Bias Disclosure
W
e acknowledge two potential sources of risk and bias in this study:
Further studies conducted by academic criminologists via public funding on the eꢀect of B W
C analytics and
professionalism training in a randomized controlled trial would reduce both of these risk s and biases.
1)
L
um, C. et al.
, “Body-worn cameras
’ eꢀects on police
oꢁcers and citizen behavior: A systematic review”
,
Campbell S
ystematic R
eviews
2020
2)
T
yler
, T
.
"L
egitimacy and Procedural Justice: A New
Element of P
olice L
eadership
"
,
2014
3)
W
ood, G. et al.
, "Procedural justice training reduces
police use of force and complaints against oꢁcers
"
,
PNA
S
2020
4)
P
eyton, K. et al.
, “
A field e
xperiment on community
policing and police legitimacy”
, PNA
S
2019
********doi.
org/10
.
1073/pnas.
1910157116
********doi.
org/10
.48550/ar
Xiv
.2009
.06732
********doi.org/10.1073/pnas.1920671117
********onlinelibrary
.
wiley
.
com/doi/10
.
1002/cl2.
1112
***********
.
ojp
.
gov/ncjrs/virtual-library/abstracts/legitim
acy-and-procedural-justice-new-element-police-leaders
hip
5) T
ay
, Y
. et al., “Eꢁcient T
ransformers: A Survey
,
”
2022
1. As B
W
C analytics are a new field with limited e
xplicit federal funding resources, funding and resources for
conducting this study
, including data processing and personnel for evaluating accuracy and results, were provided by T
ruleo
, the company producing the B
W
C analytics software that Alameda PD purchased as referenced in this study
.
2. As Alameda PD in general has a high level of professional interactions, Chief Nishant Joshi is a new chief to the Alameda P
olice Department, and the department size is around 88 oꢁcers, the overall observed eꢀects of this study may be diꢁcult to generalize
.
T
ruleo analyzes police body camera videos using artificial intelligence to help promote police professionalism. T
ruleo
work
ed with FBI National Academy alumni to build the
models that deconstruct language used during events to
help agencies promote
best practices,
train new oꢀcers,
and mitigate risk. T
o
learn
more
about
T
ruleo
'
s mission to
improve trust in the police with body camera analytics, visit ***
.
truleo
.
co
06 C
A
SE S
TUD
Y
ALAMED
A PD