Today more then ever, children are living in a hyrbid world. Cyberbullying, alongside other forms of text based harassment online is a pressing issue our app aims to reduce through an AI monitored communication service. Our product, WatchDog, aims to be "facebook messenger for kids", and seeks to empower parents to know if their child is being harassed or being sent inappropriate messages.
This prototype was produced for the UNSW WIT 2022 Hackathon, and uses sentiment analysis through the IBM Watson model to analyse the nature of the conversation your chils is having. If it detects repeated hostility or inappopriate messages, it flags it for review and sends an email to the parent with extracts of the messages, so that they can step in and decide how to proceed.
- Conversations are short - how you can be confident your sentiment analysis is on the money?
If this was an actual product, we would have a self trained model, trained on large datasets (lots of messaging data available), which would be trained to be more accurate over time. Additionally, as we only notify the parent, we empower them to exercise their judgement and decide, rather then say escalating to the police where false positives are more of a risk.
- Isn't this invasive?
This is designed to be used for younger children, and a lot of the apps marketing would be designed to enforce that. While this means the young child would lose privacy, we feel its worth the tradeoff to protect young children from cyberbullying and predators online in an increasingly virtual world.