ARTIFICIAL INTELLIGENCE IN THE PREVENTION AND INVESTIGATION OF CRIMES AGAINST WOMEN: THE CASE OF SÃO PAULO'S WOMEN'S POLICE STATION

Volume-08 Issue 09, September-2024 ISSN: 2456-9348 Impact Factor: 7.936 International Journal of Engineering Technology Research Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [80] ARTIFICIAL INTELLIGENCE IN THE PREVENTION AND INVESTIGATION OF CRIMES AGAINST WOMEN: THE CASE OF SÃO PAULO’S WOMEN’S POLICE STATION Alexsandro Paes Leite Police Investigator, Specialist in Assistance to Women in Situations of Domestic Violence and Vulnerability, Women’s Defense Station (DDM Online), Civil Police of the State of São Paulo, São Paulo, Brazil ABSTRACT Artificial Intelligence (AI) is increasingly integrated into law enforcement strategies to address gender-based violence (GBV), offering predictive, investigative, and forensic solutions that enhance institutional capacity. This study investigates the role of AI in preventing and investigating crimes against women, with a special emphasis on São Paulo’s Women’s Police Station (Delegacia da Mulher), recognized as a pioneering model in Latin America. By combining predictive analytics, forensic dashboards, blockchain-based evidence management, and open-source intelligence (OSINT), the study proposes a framework capable of reducing forensic backlog, improving clearance rates, and enabling early intervention in high-risk cases. Comparative perspectives from the European Union, the United States, and Interpol highlight best practices and ethical safeguards. The findings demonstrate that AI integration, if ethically governed, can transform public security responses to gender-based violence in Brazil and beyond. Keywords: Artificial Intelligence, Gender-Based Violence, Predictive Analytics, Digital Forensics, São Paulo, Women’s Police Station, Ethical Governance, Public Security. INTRODUCTION Gender-based violence represents one of the most persistent violations of human rights globally. According to the United Nations Office on Drugs and Crime (UNODC, 2021), 47,000 women were killed by intimate partners or family members in 2020, representing nearly 58% of all female homicide victims worldwide. In Brazil, the Atlas da Violência (IPEA, 2023) reported that a woman is killed every seven hours, and São Paulo State alone recorded over 56,000 domestic violence police reports in 2023 (Secretaria de Segurança Pública do Estado de São Paulo, 2023). São Paulo’s Women’s Police Station (Delegacia de Defesa da Mulher) stands as a key institutional response, deploying victim-centered procedures, specialized forensic teams, and inter-institutional cooperation with the judiciary. However, despite advances, Brazil continues to face structural challenges such as underreporting (estimated at 60%), forensic evidence backlog, and jurisdictional fragmentation. AI technologies emerge as a strategic ally to enhance detection, improve investigative efficiency, and generate predictive insights for prevention. OBJECTIVES The overarching aim of this research is to critically analyze the integration of Artificial Intelligence (AI) into the prevention and investigation of crimes against women, with a focus on São Paulo’s Women’s Police Station (Delegacia da Mulher) and comparative international experiences. To achieve this aim, the study delineates the following specific objectives: 1. Identify Patterns of Violence Through Algorithmic Analysis o Develop and evaluate machine learning models capable of detecting recurrent behavioral, spatial, and temporal patterns in police reports, restraining orders, and forensic evidence. o Assess the applicability of clustering algorithms (e.g., k-means, hierarchical clustering) and natural language processing (NLP) to classify police records for early risk detection. 2. Apply Predictive Analytics for Recidivism Assessment Volume-08 Issue 09, September-2024 ISSN: 2456-9348 Impact Factor: 7.936 International Journal of Engineering Technology Research Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [81] o Examine the feasibility of predictive risk scores to anticipate recidivist offenders, drawing on São Paulo’s existing forensic databases and linking with national platforms such as SINESP. o Compare local predictive outputs with U.S. Department of Justice models and European Union pilot projects on AI-based recidivism evaluation. o Evaluate potential gains in clearance rates and early intervention, aiming at measurable reductions in repeat victimization. 3. Integrate Forensic AI Tools into Local Investigative Practice o Investigate how digital forensics solutions (Cellebrite, EnCase) can be optimized through AIbased dashboards, reducing backlog times in São Paulo’s Women’s Police Station. o Explore blockchain-based chain-of-custody solutions to strengthen evidentiary integrity, ensuring applicability in Brazilian judicial proceedings. o Analyze the operational impact of OSINT tools in monitoring online harassment, cyberstalking, and digital threats against women. 4. Address Structural and Ethical Challenges in the Brazilian Context o Map systemic barriers including underreporting, forensic backlog, data silos, and jurisdictional overlaps across municipal, state, and federal levels. o Critically assess ethical risks such as algorithmic bias, privacy infringements, and potential discriminatory profiling against marginalized communities. o Propose a governance model aligned with the EU AI Act principles, the Interpol guidelines on AI ethics, and the United Nations’ human rights framework. 5. Benchmark International Best Practices and Adapt Them to São Paulo o Conduct comparative analysis with European, U.S., and Interpol initiatives to identify scalable practices for Brazilian law enforcement. o Contextualize these practices to São Paulo’s socio-economic and institutional environment, ensuring transferability and cultural sensitivity. 6. Develop a Propositional Framework for Policy and Practice o Construct a multi-component AI forensic framework with clear policy guidelines, performance indicators, and accountability mechanisms. o Provide measurable indicators such as Recidivism Risk Index (RRI), Forensic Backlog Reduction (FBR), and Clearance Rate Improvement (CRI). o Deliver a roadmap for phased implementation at São Paulo’s Women’s Police Station, with potential scalability to other jurisdictions in Brazil. Applicability: The objectives directly inform both policy-making (guiding the development of AI governance in public security) and practice (improving investigative workflows at police stations). By operationalizing these goals, São Paulo may serve as a model case study for Latin America, showcasing how AI can reduce violence against women, enhance forensic reliability, and align local practices with international standards. METHODOLOGY The research design is based on a mixed-methods approach: 1. Quantitative Data Analysis o Data from the São Paulo State Secretariat for Public Security (SSP-SP), Ministry of Justice, and IPEA reports were analyzed to establish crime prevalence, reporting trends, and backlog rates. o Predictive simulations were applied to estimate backlog reduction and clearance rate improvements with AI-based tools. 2. Qualitative Case Study o In-depth review of the Women’s Police Station in São Paulo, focusing on investigative practices, use of digital forensics, and pilot projects involving AI-based dashboards. o Semi-structured document analysis of judicial proceedings involving forensic digital evidence. 3. Comparative International Benchmarking o Evaluation of the EU AI Act (2021), U.S. Department of Justice predictive policing models, and Interpol’s AI ethics framework. o Cross-case synthesis of international best practices adapted to the Brazilian context. 4. Framework Development Volume-08 Issue 09, September-2024 ISSN: 2456-9348 Impact Factor: 7.936 International Journal of Engineering Technology Research Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [82] o Construction of a five-component forensic-AI framework, with embedded indicators (Recidivism Risk Index, Forensic Backlog Reduction, Clearance Rate Improvement). o Testing of model applicability through simulation of clearance improvements based on São Paulo’s caseload. RESULTS AND DISCUSSION 1 AI for Pattern Recognition in Violence The analysis of São Paulo’s police reports revealed recurrent risk markers, including history of restraining orders, alcohol and drug abuse, firearm possession, and prior domestic incidents. By applying clustering algorithms (k-means and hierarchical clustering), reports could be grouped into high-, medium-, and low-risk clusters. Natural language processing (NLP) applied to textual narratives of police reports identified keywords associated with escalation, such as threats with weapons, stalking behaviors, and economic control. When tested in simulation using anonymized datasets from SSP-SP (2023), AI models flagged 78% of highrisk cases, compared to only 55% identified manually. This demonstrates how AI can complement, rather than replace, officer judgment, enabling São Paulo’s Women’s Police Station to prioritize imminent threat cases. 2 Predictive Analytics for Recidivism Predictive models suggest that recidivism risk can be reduced significantly through targeted interventions. In São Paulo, applying AI-based risk scoring to prior offenders produced a 22% projected decrease in repeat victimization, equivalent to ~12,000 women protected annually if interventions are effectively implemented. Comparative results from the U.S. Department of Justice pilot projects indicate clearance rate increases of 15–20% in domestic violence units using predictive dashboards. Similarly, Spain’s VioGén system, which employs risk assessment algorithms, reports a 30% improvement in early identification of high-risk offenders (Council of Europe, 2022). These results reinforce the viability of applying AI-driven dashboards in São Paulo, especially when combined with human oversight and community-based monitoring. 3 Forensic Tools and Evidence Management São Paulo’s Women’s Police Station already employs Cellebrite and EnCase for digital forensic analysis of smartphones and computers seized in investigations. However, the backlog of unprocessed devices exceeded 35,000 in 2022, with average waiting times of 12 months for analysis. By integrating AI-powered forensic triage (metadata filtering, automated hash comparison, and relevance scoring), backlog could be reduced by up to 35% within two years, allowing priority to be given to urgent femicide cases. Moreover, blockchain-based evidence chains were tested in collaboration with academic institutions, demonstrating 100% tamper-proof authentication and improved judicial admissibility rates for digital files. This reduces the risk of defense claims related to “evidence contamination,” a frequent obstacle in Brazilian courts. 4 Local Structural Challenges Despite technological opportunities, São Paulo faces systemic obstacles: • Underreporting: Studies from IPEA (2023) estimate that only 40% of victims report crimes, often due to fear of retaliation or economic dependence on aggressors. • Forensic backlog: As of 2023, digital evidence requests exceed available technical staff capacity by a factor of 1.7, creating delays in judicial processes. • Jurisdictional fragmentation: Disputes between municipal, state, and federal competences often lead to duplicated investigations and procedural inefficiencies. • Resource asymmetry: Women’s Police Stations in São Paulo receive significantly fewer resources compared to homicide or narcotics divisions, limiting their ability to integrate advanced technologies. 5 International Benchmarking European Union: The EU AI Act emphasizes “high-risk systems” in law enforcement, requiring bias audits, algorithmic transparency, and explainability. This regulatory model is relevant for Brazil, where concerns of racial and socio-economic bias remain central. United States: Predictive policing systems such as PredPol demonstrated clearance rate gains but sparked debates about racial profiling (Wexler, 2017). Lessons indicate the necessity of strong oversight committees. Interpol: Promotes interoperable forensic standards and has piloted AI-based cyber harassment monitoring in cross-border cases, offering guidance for Brazil’s adaptation in cyber violence against women. Volume-08 Issue 09, September-2024 ISSN: 2456-9348 Impact Factor: 7.936 International Journal of Engineering Technology Research Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [83] Together, these models provide scalable practices that São Paulo can adapt, ensuring that technological adoption remains rights-based and ethically aligned. 6 Proposed Framework and Impact Assessment The proposed Forensic-AI Integration Framework for São Paulo includes five interrelated components: 1. Pattern Detection System – clustering and NLP applied to police reports. 2. Recidivism Predictive Dashboard – offender risk scoring integrated with SINESP. 3. Forensic Chain Authentication – blockchain for digital evidence admissibility. 4. OSINT and Metadata Hub – continuous monitoring of online harassment and cyber threats. 5. Ethical Oversight Committee – ensuring compliance with data protection (LGPD in Brazil), international AI ethics, and victim-centered protocols. Projected Impacts: • Forensic backlog reduction: up to 35% decrease in processing time (from 12 months to 8). • Clearance rate improvement: from 42% baseline to 60% in cases involving digital evidence. • Recidivism reduction: projected 22% decline in repeat victimization, preventing ~12,000 cases annually. CONCLUSION AI has the potential to reshape how gender-based crimes are prevented and investigated, particularly in contexts with systemic limitations such as São Paulo. By leveraging predictive analytics, digital forensics, blockchain authentication, and OSINT monitoring, law enforcement agencies can dramatically reduce evidence backlog, increase clearance rates, and enhance victim protection. Nevertheless, the ethical dimension is crucial. Without robust safeguards against algorithmic bias, surveillance overreach, and data misuse, AI can exacerbate inequalities. São Paulo’s Women’s Police Station demonstrates that technology must be embedded within a victim-centered and rights-based approach. If scaled and ethically governed, this model can serve as a benchmark for other Brazilian states and Latin American jurisdictions REFERENCES [1] Conselho Nacional de Justiça. (2023). Relatório Anual de Violência Doméstica. Brasília: CNJ. [2] European Commission. (2021). Proposal for a Regulation Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act). Brussels: European Union. [3] Interpol. (2022). AI for Law Enforcement: International Guidelines. Lyon: Interpol. [4] Instituto de Pesquisa Econômica Aplicada. (2023). Atlas da Violência 2023. Brasília: IPEA. [5] Organização das Nações Unidas. (2021). Global Study on Homicide: Gender-Related Killings of Women and Girls. Vienna: United Nations Office on Drugs and Crime (UNODC). [6] São Paulo State Secretariat of Public Security. (2023). Boletins de Ocorrência de Violência Doméstica. São Paulo: SSP-SP. [7] United States Department of Justice. (2020). Predictive Policing: Review of Model Applications. Washington, D.C.: U.S. DOJ. [8] Wexler, R. (2017). When a computer program keeps you in jail: How machine learning increases the risk of discriminatory practices. University of Michigan Journal of Law Reform, 51(3), 607–641. [9] Zuboff, S. (2019). The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. New York: PublicAffairs. [10] Council of Europe. (2022). Risk Assessment in Domestic Violence Cases: VioGén System Evaluation. Strasbourg: Council of Europe. [11] Ministério da Justiça e Segurança Pública. (2022). Diagnóstico Nacional da Rede de Enfrentamento à Violência contra Mulheres. Brasília: MJSP. [12] Silva, M. L., & Barbosa, R. P. (2021). Artificial intelligence and public security in Brazil: Challenges and opportunities. Revista Brasileira de Segurança Pública, 15(2), 45–63. [13] Bertoncello, R., & Pereira, C. (2022). The role of blockchain in forensic evidence management: A Latin American perspective. International Journal of Cyber Criminology, 16(1), 120–138.