Business Challenge
Verbal bullying is often dismissed as “just words,” despite being one of the most harmful and common forms of bullying. Schools focus on physical safety, while emotional harm remains invisible. Although Indian schools are equipped with extensive CCTV systems, these only capture visual data and cannot interpret behaviour, tone or intent. Counselling systems are often standardised, failing to address different bullying patterns.
)
Human Perspective
Students experience repeated verbal bullying around appearance, identity, language or social exclusion, which becomes internalised over time. Most incidents go unreported because they don’t escalate into visible conflict, blending into everyday classroom noise. This leaves students unsupported and counsellors unaware, despite harm happening in plain sight.
)
Social need
Schools are mandated to install CCTV to ensure safety, but current systems only address visible misconduct. There is a broader need to recognise emotional well-being as part of student safety. By identifying behavioural patterns instead of individuals, schools can shift from punitive responses to empathy-led support, addressing a global issue that has long remained unseen.
)
Lab process
The team found verbal abuse is rarely detected or reported. This led to the idea of using unused CCTV audio feeds. AI was developed to analyse tone, conversational intensity and repeated patterns linked to bullying. The system was built with privacy-first principles: no recording storage, no identification of individuals, and fully anonymised outputs. Insights are aggregated into dashboards, helping principals and counsellors.
Innovation outcome
Bullying Decoder transforms classroom sound into a real-time behavioural insight system. By decoding tone, repetition and context, not just words, it identifies patterns of verbal bullying and translates them into actionable insights. The system redefines AI as a tool for care, enabling early intervention while respecting anonymity.
Our work / Bullying Decoder 001 / 000