I-Corps: Exploring the Commercialization Potential of ChatCoder

Ursinus College, Collegeville PA

Investigators

Abstract

Current parental control software does not provide parents and kids with the one tool they desperately want and need ? a method for halting cyber-aggression in progress, rather than reporting the event after it occurs. Although most cyberbullying and cyber-predation acts occur over an extended period of time, current programs can only detect acts based on keywords in text; they don?t offer tools to actually stop the abuse and they don?t capture context. Concern about cyberbullying and cyber-predation among parents and authorities has led to a booming parental control software market and to the passage of anti-cyberbullying laws in many states. The proposed innovation offers better recognition of cyber-abuse and as well as response capabilities; thereby providing a more dynamic, interactive, and empowering resource for youths and their parents. The team has developed machine learning algorithms that are able to detect approximately 80% of cyberbullying communication, and 84% of predatory conversation. These algorithms were developed using data that was collected and labeled for research purposes. The algorithms need to be adapted for real-time communication, as well as continuously updated as additional data become available. The theoretical model for victim response is in development. The model identifies the type, context and severity of bullying or predation, along with categories of potential responses from which victims can choose. The team has completed the proof-of-concept and is ready to move forward with the development of a software prototype.