Safety by Design – Deterring a New Kind of Online Abuse

  • 2023

  • Social Impact

Abuse in transaction descriptions (AITD) are a new insidious form of online abuse sent through bank transactions. Currently unblockable and intimately tied to one’s day-to-day life, CommBank needed immediate and long-lasting solutions to make banking safer. Validated with customers, Commbank’s Community Wellbeing Team frontline staff and seven safety advocacy organisations.

  • MORE
  • As the New Payments Platform (NPP) increases the length of bank transaction descriptions, perpetrators are subverting this feature to send unblockable messages to harass, intimidate and abuse. Intertwined with finance and daily lives, the impact is immense. This abuse is becoming more commonplace and at higher severity. Working with data scientists at CommBank, we determined 582 unique senders of abusive transactions in July 2021, with a 193% y.o.y increase in worst of the worst abusive transactions. With future NPP features like image and document attachments upcoming, the bandwidth for abuse continues to widen.

  • We developed a safe intervention strategy to identify and assess abusive transactions, alongside 15 prototypes of long-term future processes to deliver immediate and sustainable impact: Protect: Solutions to immediately reduce the volume of abuse and prevent uptake of new senders. Support: Pathways to access help privately and safely with the Commbank Community Wellbeing Team. Enable: Connect customers to a network of external specialists, minimising where customers need to retell their story. Designed in collaboration with support organisations and customers to ensure that interventions were safe, carried out with consent and fulfilled the role that customers expect their bank to play.

  • Early on, we realised that the definition of abuse is far greater and more nuanced than simply blocking profanity as Commbank’s previous word filter did. We developed a new definition that includes harassment, hate speech, stalking, coercive control, threats and intimidation, humiliation and condescension, manipulation and unwanted sexual requests; in addition to offensive language intended to control and cause harm. We worked closely with Commbank’s AI Labs to embed this within their machine learning technology. We developed other indicators such as frequency, cadence, monetary amount, relationship to the sender and more as factors to flag, for humans to act on.

  • Supported by a human-centred approach, backed by a rigorous testing methodology: - 28+ Hours of qualitative testing with specialist services and customers with lived experience. - 20+ Hours of subject matter expert interviews & co-design - 18 Unique intervention prototypes tested - 5 rounds of user testing This research revealed many important and unintuitive learnings. From WAGEC and WESNET in particular, we learned that any support mechanism can be weaponised by abusers to create further harm, and must be designed extremely carefully with that lens in mind. Secondly, simple solutions require additional consideration in financial settings. Recipients’ financial situations are often interwoven with their abuser, meaning interventions, like sending a warning letter, can have severe and dangerous consequences. “One of the most important insights this project revealed was how critical it was to have the safe consent of a customer for us to intervene - it’s not as simple as having a zero tolerance stance.” - Caroline Wall, Commbank project sponsor Similarly, further complexity comes when the perpetrator is not a CommBank customer. New OFI (Other Financial Institution) pathways have been established to ensure there is a course of action for perpetrators regardless of who they bank with.