[BtB | PO Diary] Wk 3: The Quest for Personalized Support
Welcome back, curious minds! Week four of our AI chatbot testing journey has been a deep dive into a crucial aspect of protection work: location. It turns out, just like in real estate, location is everything when it comes to providing accurate and helpful information to refugees and displaced people.
The Location Puzzle:
This week, we put our AI chatbots through their paces, testing their ability to understand and respond to user location. We discovered that some bots, like Directus GPT, have started asking follow-up questions to clarify location, a promising development. However, the responses we received still need refinement. In some cases, the bots provided information for Athens or Chios when the user's location wasn't specified. This isn't ideal when dealing with hyper-local services and resources.
Integrating Processes and Services:
We also grappled with the challenge of providing holistic, client-centered responses that combine information about both processes (like how to apply for asylum) and specific services (like legal aid organizations). It became clear that while some bots excel at providing process information, others shine at identifying services. Finding the perfect balance is key to delivering truly effective support.
Bugs, Glitches, and Gremlins:
Oh, the joys of technology! This week wasn't without its technical hiccups. We encountered bugs in both our testing environment and the Refugee.Info website, which made cross-checking references a bit of a headache. However, these challenges are all part of the process and offer valuable learning opportunities as we refine our AI tools.
Lessons Learned (So Far):
Location matters: Understanding a user's location is crucial for providing accurate and relevant information. We need to improve our bots' ability to factor in location when generating responses.
The whole picture: Integrating information about both processes and services is essential for delivering comprehensive support. We're exploring ways to bridge this gap and provide more holistic responses.
Bugs are our friends: Technical challenges are inevitable, but they also help us identify areas for improvement and build more robust AI systems.
Looking Ahead:
As we continue our AI testing journey, we're focusing on refining location-based responses, integrating processes and services, and addressing any bugs or glitches that arise. We're also excited to explore new avenues, such as using system prompts to encourage the bots to ask follow-up questions about legal status and gender, which could further enhance the personalization of their responses.
We hope you're enjoying this behind-the-scenes look at our work. Stay tuned for our next post as we continue to push the boundaries of AI and its potential to empower refugees and displaced people worldwide.
#TechForGood #HumanitarianAI #AIWithHeart #DigitalProtection
