Tell Polis

Background

The National Anti-Violence and Sexual Harassment Helpline (NAVH) received 8,400 calls in 2021, which increased to 10,800 calls in 2022. These increases may reflect increased awareness and therefore increased reporting, yet they also highlight the persistent prevalence of domestic violence in Singapore and the ongoing need for timely intervention to provide help to vulnerable individuals.

Current Referral & Triaging Process

We use the child abuse reporting and triaging process to illustrate the current referral and triaging process. This is broadly similar to other types of domestic violence cases.

Today, professionals in the community (e.g. social service agency staff) are trained to use the localised Structured Decision Making (SDM) system - a set of evidence-based decision-making tools that guide them on managing reports of suspected abuse – to decide on appropriate interventions for survivors.

The escalation pathway is as follows:

1. Frontline professionals, such as social workers in Family Service Centres (FSCs), use the Sector-Specific Screening Guide (SSSG) and Child Abuse Reporting Guide (CARG) to decide how to manage reports of suspected child abuse, and the follow-up thereafter to ensure the safety and well-being of the child(ren).
   
   Child protection concerns operate on a continuum, with intervention from either the community, more specialised partners or MSF's Child Protective Services (CPS). Generally, a more serious concern requires statutory intervention with CPS stepping in when warranted.
   

2. Most frontline social workers currently spend at least three hours per case solely on writing case notes and assessing how to manage reports of suspected child abuse using the SSSG and/orCARG. This is because information required for a case report and subsequent assessment on its severity comes from fragmented sources, which the social worker can only collect during interactions with clients.
   
   When reports are escalated to the NAVH, while urgent cases are acted upon immediately, there may be a need for NAVH to seek clarifications and obtain more information and/or comprehensive assessment from relevant parties for non-urgent cases.

Problem Statement

Our team wondered how we might further improve the reporting and triaging process for survivors of domestic violence to provide better support for this vulnerable group.

Pain point 1

Different frameworks and criteria are used to assess how serious cases are, and these can be interpreted in different ways.

Pain point 2

Deciding on suitable actions for serious cases can take a long time due to back and forth clarifications, which may delay timely help.

Hypothesis

With the increase in the number of suspected abuse cases being reported, enhancing the reporting process can lead to more efficient resource allocation, quicker response times, and better support for survivors by social service agencies, which include Family Service Centres, specialist agencies, and MSF. From the survivor's perspective, a streamlined and effective reporting system would encourage more survivors to come forward if there is assurance that cases are addressed promptly and effectively. This also avoids subjecting victims to re-traumatisation when they are requested to recount their traumatic experience multiple times to various assessing parties.

Objectives & Product Goals

Our team is building an AI tool that helps social workers in FSCs parse case notes into concise case summaries and identify missing information that may be necessary for reporting. The expected impact of an effective tool would be:

• Improved workflow efficiency (i.e. reduction in back and forth between social workers, NAVH, and MSF to seek clarifications and obtain more information)

• Improved decision-making quality (i.e. increase in percentage of referrals to NAVH that get accepted), including being able to identify potential cases early on for timely intervention

Solutions Overview

Key Features

The AI-powered tool parses social workers’ raw case notes to generate three outputs:

• Inputs for the referral form containing salient points of the social worker’s key concerns about the case

• 1-page concise summary of the case

• Identified missing information that is necessary for reporting

Research & Approach

Problem Discovery

Our team conducted user interviews with survivors of domestic violence, people who have had close experience with a loved one or friend who had experienced domestic violence, as well as various stakeholders across social service agencies; social workers in FSCs, school counsellors, police officers, NAVH officers, and CPS officers.

Our goal was to:

• Understand the existing reporting and triaging process for domestic violence cases from not just the victim’s perspective, but also that of social workers and other officers involved in managing the case

• Understand the pain points and challenges of officers managing the case, and what their key objectives and considerations were

• Gather insights on potential opportunities where tech could be an enabler for positive impact or reduce pain points

Scribe is an existing product that already served the user need of social workers looking for a tool to reduce the amount of time they spent writing case notes and reports. It does so by synthesising social workers’ interactions into meaningful summaries based on recorded audio conversations. We then tried to seek opportunities to create impact in other ways.

Pilot

Our team will be conducting pilots with six social workers from two FSCs who directly handle domestic violence cases, and their supervisors who vet their case reports.

MSF (NAVH and CPS/APS) had also surfaced concerns about the risks associated with AI-generated outputs, and discussed how these risks could be mitigated or treated. They will also be participating in the pilot to give feedback on how closely the content generated aligned with actual ones.

Next Steps & Future Vision

In the next phase, we hope to pilot this tool with more FSCs, NAVH and MSF to improve its effectiveness and accuracy. This tool can be integrated into Social360, a case management system accessible by MSF and all SSAs, as an additional feature in the future. There is potential for an end-to-end workflow to be streamlined by transcribing raw case notes from audio recordings, then automatically parsing it through our AI-powered evaluation tool to generate referral reports and 1-page summaries.

Domestic violence is no private matter. Your intervention can help save lives. If you are willing to share any experience or suggestions that might help us improve the reporting and triaging process for domestic violence, please reach out to the team.

Team

Tell Polis was built by a multi-disciplinary team of engineers, product manager, product operations and policy officers.

We also thank Toby Gail for her contributions to the project in the research phase.