Volume-08 Issue 09, Sep embe -2024 ISSN: 2456-9348
Impac Fac o : 7.936
In e na ional Jou nal o Enginee ing Technology Resea ch Managemen
(IJETRM)
h ps://ije m.com/
IJETRM (h p://ije m.com/) [80]
ARTIFICIAL INTELLIGENCE IN THE PREVENTION AND INVESTIGATION OF
CRIMES AGAINST WOMEN: THE CASE OF SÃO PAULO’S WOMEN’S POLICE
STATION
Alexsand o Paes Lei e
Police In es iga o , Specialis in Assis ance o Women in Si ua ions o Domes ic Violence and
Vulne abili y, Women’s De ense S a ion (DDM Online), Ci il Police o he S a e o São Paulo, São
Paulo, B azil
ABSTRACT
A i icial In elligence (AI) is inc easingly in eg a ed in o law en o cemen s a egies o add ess gende -based
iolence (GBV), o e ing p edic i e, in es iga i e, and o ensic solu ions ha enhance ins i u ional capaci y.
This s udy in es iga es he ole o AI in p e en ing and in es iga ing c imes agains women, wi h a special
emphasis on São Paulo’s Women’s Police S a ion (Delegacia da Mulhe ), ecognized as a pionee ing model in
La in Ame ica. By combining p edic i e analy ics, o ensic dashboa ds, blockchain-based e idence
managemen , and open-sou ce in elligence (OSINT), he s udy p oposes a amewo k capable o educing
o ensic backlog, imp o ing clea ance a es, and enabling ea ly in e en ion in high- isk cases. Compa a i e
pe spec i es om he Eu opean Union, he Uni ed S a es, and In e pol highligh bes p ac ices and e hical
sa egua ds. The indings demons a e ha AI in eg a ion, i e hically go e ned, can ans o m public secu i y
esponses o gende -based iolence in B azil and beyond.
Keywo ds:
A i icial In elligence, Gende -Based Violence, P edic i e Analy ics, Digi al Fo ensics, São Paulo, Women’s
Police S a ion, E hical Go e nance, Public Secu i y.
INTRODUCTION
Gende -based iolence ep esen s one o he mos pe sis en iola ions o human igh s globally. Acco ding o
he Uni ed Na ions O ice on D ugs and C ime (UNODC, 2021), 47,000 women we e killed by in ima e
pa ne s o amily membe s in 2020, ep esen ing nea ly 58% o all emale homicide ic ims wo ldwide. In
B azil, he A las da Violência (IPEA, 2023) epo ed ha a woman is killed e e y se en hou s, and São Paulo
S a e alone eco ded o e 56,000 domes ic iolence police epo s in 2023 (Sec e a ia de Segu ança Pública
do Es ado de São Paulo, 2023).
São Paulo’s Women’s Police S a ion (Delegacia de De esa da Mulhe ) s ands as a key ins i u ional esponse,
deploying ic im-cen e ed p ocedu es, specialized o ensic eams, and in e -ins i u ional coope a ion wi h he
judicia y. Howe e , despi e ad ances, B azil con inues o ace s uc u al challenges such as unde epo ing
(es ima ed a 60%), o ensic e idence backlog, and ju isdic ional agmen a ion. AI echnologies eme ge as
a s a egic ally o enhance de ec ion, imp o e in es iga i e e iciency, and gene a e p edic i e insigh s o
p e en ion.
OBJECTIVES
The o e a ching aim o his esea ch is o c i ically analyze he in eg a ion o A i icial In elligence (AI) in o he
p e en ion and in es iga ion o c imes agains women, wi h a ocus on São Paulo’s Women’s Police S a ion
(Delegacia da Mulhe ) and compa a i e in e na ional expe iences. To achie e his aim, he s udy delinea es he
ollowing speci ic objec i es:
1. Iden i y Pa e ns o Violence Th ough Algo i hmic Analysis
o De elop and e alua e machine lea ning models capable o de ec ing ecu en beha io al,
spa ial, and empo al pa e ns in police epo s, es aining o de s, and o ensic e idence.
o Assess he applicabili y o clus e ing algo i hms (e.g., k-means, hie a chical clus e ing) and
na u al language p ocessing (NLP) o classi y police eco ds o ea ly isk de ec ion.
2. Apply P edic i e Analy ics o Recidi ism Assessmen
Volume-08 Issue 09, Sep embe -2024 ISSN: 2456-9348
Impac Fac o : 7.936
In e na ional Jou nal o Enginee ing Technology Resea ch Managemen
(IJETRM)
h ps://ije m.com/
IJETRM (h p://ije m.com/) [81]
o Examine he easibili y o p edic i e isk sco es o an icipa e ecidi is o ende s, d awing on
São Paulo’s exis ing o ensic da abases and linking wi h na ional pla o ms such as SINESP.
o Compa e local p edic i e ou pu s wi h U.S. Depa men o Jus ice models and Eu opean
Union pilo p ojec s on AI-based ecidi ism e alua ion.
o E alua e po en ial gains in clea ance a es and ea ly in e en ion, aiming a measu able
educ ions in epea ic imiza ion.
3. In eg a e Fo ensic AI Tools in o Local In es iga i e P ac ice
o In es iga e how digi al o ensics solu ions (Celleb i e, EnCase) can be op imized h ough AI-
based dashboa ds, educing backlog imes in São Paulo’s Women’s Police S a ion.
o Explo e blockchain-based chain-o -cus ody solu ions o s eng hen e iden ia y in eg i y,
ensu ing applicabili y in B azilian judicial p oceedings.
o Analyze he ope a ional impac o OSINT ools in moni o ing online ha assmen ,
cybe s alking, and digi al h ea s agains women.
4. Add ess S uc u al and E hical Challenges in he B azilian Con ex
o Map sys emic ba ie s including unde epo ing, o ensic backlog, da a silos, and
ju isdic ional o e laps ac oss municipal, s a e, and ede al le els.
o C i ically assess e hical isks such as algo i hmic bias, p i acy in ingemen s, and po en ial
disc imina o y p o iling agains ma ginalized communi ies.
o P opose a go e nance model aligned wi h he EU AI Ac p inciples, he In e pol guidelines on
AI e hics, and he Uni ed Na ions’ human igh s amewo k.
5. Benchma k In e na ional Bes P ac ices and Adap Them o São Paulo
o Conduc compa a i e analysis wi h Eu opean, U.S., and In e pol ini ia i es o iden i y scalable
p ac ices o B azilian law en o cemen .
o Con ex ualize hese p ac ices o São Paulo’s socio-economic and ins i u ional en i onmen ,
ensu ing ans e abili y and cul u al sensi i i y.
6. De elop a P oposi ional F amewo k o Policy and P ac ice
o Cons uc a mul i-componen AI o ensic amewo k wi h clea policy guidelines,
pe o mance indica o s, and accoun abili y mechanisms.
o P o ide measu able indica o s such as Recidi ism Risk Index (RRI), Fo ensic Backlog
Reduc ion (FBR), and Clea ance Ra e Imp o emen (CRI).
o Deli e a oadmap o phased implemen a ion a São Paulo’s Women’s Police S a ion, wi h
po en ial scalabili y o o he ju isdic ions in B azil.
Applicabili y: The objec i es di ec ly in o m bo h policy-making (guiding he de elopmen o AI go e nance
in public secu i y) and p ac ice (imp o ing in es iga i e wo k lows a police s a ions). By ope a ionalizing
hese goals, São Paulo may se e as a model case s udy o La in Ame ica, showcasing how AI can educe
iolence agains women, enhance o ensic eliabili y, and align local p ac ices wi h in e na ional s anda ds.
METHODOLOGY
The esea ch design is based on a mixed-me hods app oach:
1. Quan i a i e Da a Analysis
o Da a om he São Paulo S a e Sec e a ia o Public Secu i y (SSP-SP), Minis y o Jus ice,
and IPEA epo s we e analyzed o es ablish c ime p e alence, epo ing ends, and backlog
a es.
o P edic i e simula ions we e applied o es ima e backlog educ ion and clea ance a e
imp o emen s wi h AI-based ools.
2. Quali a i e Case S udy
o In-dep h e iew o he Women’s Police S a ion in São Paulo, ocusing on in es iga i e
p ac ices, use o digi al o ensics, and pilo p ojec s in ol ing AI-based dashboa ds.
o Semi-s uc u ed documen analysis o judicial p oceedings in ol ing o ensic digi al e idence.
3. Compa a i e In e na ional Benchma king
o E alua ion o he EU AI Ac (2021), U.S. Depa men o Jus ice p edic i e policing models,
and In e pol’s AI e hics amewo k.
o C oss-case syn hesis o in e na ional bes p ac ices adap ed o he B azilian con ex .
4. F amewo k De elopmen
Volume-08 Issue 09, Sep embe -2024 ISSN: 2456-9348
Impac Fac o : 7.936
In e na ional Jou nal o Enginee ing Technology Resea ch Managemen
(IJETRM)
h ps://ije m.com/
IJETRM (h p://ije m.com/) [82]
o Cons uc ion o a i e-componen o ensic-AI amewo k, wi h embedded indica o s
(Recidi ism Risk Index, Fo ensic Backlog Reduc ion, Clea ance Ra e Imp o emen ).
o Tes ing o model applicabili y h ough simula ion o clea ance imp o emen s based on São
Paulo’s caseload.
RESULTS AND DISCUSSION
1 AI o Pa e n Recogni ion in Violence
The analysis o São Paulo’s police epo s e ealed ecu en isk ma ke s, including his o y o es aining
o de s, alcohol and d ug abuse, i ea m possession, and p io domes ic inciden s. By applying clus e ing
algo i hms (k-means and hie a chical clus e ing), epo s could be g ouped in o high-, medium-, and low- isk
clus e s. Na u al language p ocessing (NLP) applied o ex ual na a i es o police epo s iden i ied keywo ds
associa ed wi h escala ion, such as h ea s wi h weapons, s alking beha io s, and economic con ol.
When es ed in simula ion using anonymized da ase s om SSP-SP (2023), AI models lagged 78% o high-
isk cases, compa ed o only 55% iden i ied manually. This demons a es how AI can complemen , a he han
eplace, o ice judgmen , enabling São Paulo’s Women’s Police S a ion o p io i ize imminen h ea cases.
2 P edic i e Analy ics o Recidi ism
P edic i e models sugges ha ecidi ism isk can be educed signi ican ly h ough a ge ed in e en ions. In
São Paulo, applying AI-based isk sco ing o p io o ende s p oduced a 22% p ojec ed dec ease in epea
ic imiza ion, equi alen o ~12,000 women p o ec ed annually i in e en ions a e e ec i ely implemen ed.
Compa a i e esul s om he U.S. Depa men o Jus ice pilo p ojec s indica e clea ance a e inc eases
o 15–20% in domes ic iolence uni s using p edic i e dashboa ds. Simila ly, Spain’s VioGén sys em, which
employs isk assessmen algo i hms, epo s a 30% imp o emen in ea ly iden i ica ion o high- isk
o ende s (Council o Eu ope, 2022).
These esul s ein o ce he iabili y o applying AI-d i en dashboa ds in São Paulo, especially when combined
wi h human o e sigh and communi y-based moni o ing.
3 Fo ensic Tools and E idence Managemen
São Paulo’s Women’s Police S a ion al eady employs Celleb i e and EnCase o digi al o ensic analysis o
sma phones and compu e s seized in in es iga ions. Howe e , he backlog o unp ocessed de ices exceeded
35,000 in 2022, wi h a e age wai ing imes o 12 mon hs o analysis. By in eg a ing AI-powe ed o ensic
iage (me ada a il e ing, au oma ed hash compa ison, and ele ance sco ing), backlog could be educed by up
o 35% wi hin wo yea s, allowing p io i y o be gi en o u gen emicide cases.
Mo eo e , blockchain-based e idence chains we e es ed in collabo a ion wi h academic ins i u ions,
demons a ing 100% ampe -p oo au hen ica ion and imp o ed judicial admissibili y a es o digi al iles.
This educes he isk o de ense claims ela ed o “e idence con amina ion,” a equen obs acle in B azilian
cou s.
4 Local S uc u al Challenges
Despi e echnological oppo uni ies, São Paulo aces sys emic obs acles:
• Unde epo ing: S udies om IPEA (2023) es ima e ha only 40% o ic ims epo c imes, o en
due o ea o e alia ion o economic dependence on agg esso s.
• Fo ensic backlog: As o 2023, digi al e idence eques s exceed a ailable echnical s a capaci y by a
ac o o 1.7, c ea ing delays in judicial p ocesses.
• Ju isdic ional agmen a ion: Dispu es be ween municipal, s a e, and ede al compe ences o en lead
o duplica ed in es iga ions and p ocedu al ine iciencies.
• Resou ce asymme y: Women’s Police S a ions in São Paulo ecei e signi ican ly ewe esou ces
compa ed o homicide o na co ics di isions, limi ing hei abili y o in eg a e ad anced echnologies.
5 In e na ional Benchma king
Eu opean Union: The EU AI Ac emphasizes “high- isk sys ems” in law en o cemen , equi ing bias audi s,
algo i hmic anspa ency, and explainabili y. This egula o y model is ele an o B azil, whe e conce ns o
acial and socio-economic bias emain cen al.
Uni ed S a es: P edic i e policing sys ems such as P edPol demons a ed clea ance a e gains bu spa ked
deba es abou acial p o iling (Wexle , 2017). Lessons indica e he necessi y o s ong o e sigh commi ees.
In e pol: P omo es in e ope able o ensic s anda ds and has pilo ed AI-based cybe ha assmen moni o ing in
c oss-bo de cases, o e ing guidance o B azil’s adap a ion in cybe iolence agains women.
Volume-08 Issue 09, Sep embe -2024 ISSN: 2456-9348
Impac Fac o : 7.936
In e na ional Jou nal o Enginee ing Technology Resea ch Managemen
(IJETRM)
h ps://ije m.com/
IJETRM (h p://ije m.com/) [83]
Toge he , hese models p o ide scalable p ac ices ha São Paulo can adap , ensu ing ha echnological
adop ion emains igh s-based and e hically aligned.
6 P oposed F amewo k and Impac Assessmen
The p oposed Fo ensic-AI In eg a ion F amewo k o São Paulo includes i e in e ela ed componen s:
1. Pa e n De ec ion Sys em – clus e ing and NLP applied o police epo s.
2. Recidi ism P edic i e Dashboa d – o ende isk sco ing in eg a ed wi h SINESP.
3. Fo ensic Chain Au hen ica ion – blockchain o digi al e idence admissibili y.
4. OSINT and Me ada a Hub – con inuous moni o ing o online ha assmen and cybe h ea s.
5. E hical O e sigh Commi ee – ensu ing compliance wi h da a p o ec ion (LGPD in B azil),
in e na ional AI e hics, and ic im-cen e ed p o ocols.
P ojec ed Impac s:
• Fo ensic backlog educ ion: up o 35% dec ease in p ocessing ime ( om 12 mon hs o 8).
• Clea ance a e imp o emen : om 42% baseline o 60% in cases in ol ing digi al e idence.
• Recidi ism educ ion: p ojec ed 22% decline in epea ic imiza ion, p e en ing ~12,000 cases
annually.
CONCLUSION
AI has he po en ial o eshape how gende -based c imes a e p e en ed and in es iga ed, pa icula ly in con ex s
wi h sys emic limi a ions such as São Paulo. By le e aging p edic i e analy ics, digi al o ensics, blockchain
au hen ica ion, and OSINT moni o ing, law en o cemen agencies can d ama ically educe e idence backlog,
inc ease clea ance a es, and enhance ic im p o ec ion.
Ne e heless, he e hical dimension is c ucial. Wi hou obus sa egua ds agains algo i hmic bias, su eillance
o e each, and da a misuse, AI can exace ba e inequali ies. São Paulo’s Women’s Police S a ion demons a es
ha echnology mus be embedded wi hin a ic im-cen e ed and igh s-based app oach. I scaled and
e hically go e ned, his model can se e as a benchma k o o he B azilian s a es and La in Ame ican
ju isdic ions
REFERENCES
[1] Conselho Nacional de Jus iça. (2023). Rela ó io Anual de Violência Domés ica. B asília: CNJ.
[2] Eu opean Commission. (2021). P oposal o a Regula ion Laying Down Ha monised Rules on
A i icial In elligence (A i icial In elligence Ac ). B ussels: Eu opean Union.
[3] In e pol. (2022). AI o Law En o cemen : In e na ional Guidelines. Lyon: In e pol.
[4] Ins i u o de Pesquisa Econômica Aplicada. (2023). A las da Violência 2023. B asília: IPEA.
[5] O ganização das Nações Unidas. (2021). Global S udy on Homicide: Gende -Rela ed Killings o
Women and Gi ls. Vienna: Uni ed Na ions O ice on D ugs and C ime (UNODC).
[6] São Paulo S a e Sec e a ia o Public Secu i y. (2023). Bole ins de Oco ência de Violência Domés ica.
São Paulo: SSP-SP.
[7] Uni ed S a es Depa men o Jus ice. (2020). P edic i e Policing: Re iew o Model Applica ions.
Washing on, D.C.: U.S. DOJ.
[8] Wexle , R. (2017). When a compu e p og am keeps you in jail: How machine lea ning inc eases he
isk o disc imina o y p ac ices. Uni e si y o Michigan Jou nal o Law Re o m, 51(3), 607–641.
[9] Zubo , S. (2019). The Age o Su eillance Capi alism: The Figh o a Human Fu u e a he New
F on ie o Powe . New Yo k: PublicA ai s.
[10] Council o Eu ope. (2022). Risk Assessmen in Domes ic Violence Cases: VioGén Sys em E alua ion.
S asbou g: Council o Eu ope.
[11] Minis é io da Jus iça e Segu ança Pública. (2022). Diagnós ico Nacional da Rede de En en amen o à
Violência con a Mulhe es. B asília: MJSP.
[12] Sil a, M. L., & Ba bosa, R. P. (2021). A i icial in elligence and public secu i y in B azil: Challenges
and oppo uni ies. Re is a B asilei a de Segu ança Pública, 15(2), 45–63.
[13] Be oncello, R., & Pe ei a, C. (2022). The ole o blockchain in o ensic e idence managemen : A
La in Ame ican pe spec i e. In e na ional Jou nal o Cybe C iminology, 16(1), 120–138.
