Responsible use of AI

Why AI? 

At Alegra Learning, we want to use technology in ways that transform students’ learning. At the same time, we are committed to being vigilant about student safety, privacy, and data security for our Joy School English users. 

Harnessing the Power of AI for Language Learning

We are committed to providing innovative tools and resources to enhance language learning. Artificial Intelligence (AI) offers a unique and powerful way to support learners in achieving their language goals. Here’s how AI can be a valuable resource in language education:

• Personalized Learning Experience
  AI enables personalized learning by adapting to each student’s pace and style. Whether you're practicing vocabulary, grammar, or pronunciation, AI can provide tailored exercises and feedback to match your students’ individual needs and help them progress effectively.
• Instant Feedback and Support
  AI tools offer instant feedback, allowing learners to correct mistakes and improve their skills in real-time. This immediate response helps reinforce learning and boosts confidence, providing a supportive environment for practice and exploration.
• Expanding Access to Resource
  With AI, language learning resources are more accessible than ever. Students can access a wealth of information and practice opportunities anytime, anywhere. This flexibility empowers learners to tailor their education to their individual needs and pace.

Privacy and Data Security

We understand that privacy and data security are of utmost importance to our users. At Alegra Learning, we prioritize protecting your personal information and ensuring a safe learning environment.

No Recording Storage

OpenAI, the provider of our AI language learning tools, does not save any student recordings to feed into large language models (LLMs). Student voice recordings and interactions with AI are used solely to provide immediate feedback and are not stored or used for any other purpose by OpenAI. 

Student recordings are saved in the password protected Joy School English Teacher Portal so that teachers and parents can listen to the recordings in order to evaluate student progress and inform instruction. 

Commitment to Confidentiality

We are dedicated to maintaining the confidentiality of your data and do not sell student or teacher data. Our systems are designed to protect your privacy and ensure that the personal information of you and your students remains secure. We continually update our security measures to safeguard against any potential risks. For further information, please refer to our Privacy Policy.

Transparency and Trust

Building trust with our users is a top priority for us. We are committed to being transparent about our use of AI and data handling practices. If you have any questions or concerns regarding our privacy practices, please feel free to contact our support team.

Safety

Alegra Learning’s latest AI-enabled activity—Kady Chain—leverages the API from OpenAI, the team that pioneered the LLM (large language model) known as ChatGPT. Joy School English users are the beneficiaries of extensive guardrails pre-built into the chatbot, including:

• Content Filtering
  ChatGPT has been trained on a wide range of text, but it includes filters and content moderation systems that help detect and avoid generating harmful, offensive, or inappropriate content. If a query or response is flagged as potentially problematic, the system is designed to avoid it or provide a neutral, non-offensive answer.
• Ethical Guidelines
  The training of ChatGPT includes ethical guidelines that discourage the generation of content that is violent, sexually explicit, hateful, or otherwise harmful. The model is designed to prioritize respect, inclusivity, and appropriateness in all interactions.
• User Reports and Feedback
  If an inappropriate response does slip through, users can report it, and such feedback is used to improve future versions of the model. This helps in continuously refining the system's safety measures.
• Prompt Reframing
  The model is designed to reframe or redirect conversations that may veer into sensitive or inappropriate topics. Instead of directly addressing certain requests, it may steer the conversation in a more constructive direction.
• Inherent Safeguards
  During the training process, data containing offensive or inappropriate content is minimized to reduce the likelihood of the model producing such responses. Additionally, ongoing updates and improvements to the model aim to enhance its ability to handle sensitive topics responsibly.

In addition to the above, Alegra Learning filters the content even more in Kady Chain, by providing guidelines to the AI bot which it then adheres to in its conversations with students. At present, we instruct it to speak as it would to a 5 year old, using simple language, short sentences, and short responses. This demographic-directed concision makes it much more likely that Kady’s responses aren’t able to stray into topical areas of impropriety or objectionable content. And by studying and observing how it is being used, we plan to continually adjust those restrictions and guidelines by adding to them and widening the filtering net to be even more effective.