Teachers and community members rightly ask questions about the environmental and ethical impact of AI. One participant left the project half way through, as he couldn't accept the use of AI in this context. His comments were very helpful to the remaining 12 in terms of increasing their awareness and responsible use.
This section lays out the issues transparently and explains how we aim to use AI in a small, careful and human-centred way.
Our approach is simple:
Less AI, not more
Short, targeted queries, not long conversations
Teacher judgement first; AI quietly in the background
Respect for Cornish, the learner, and the planet
Below is an overview of the ethical principles and sustainability practices that guide our work.
1. Human-centred learning
AI supports teachers; it never directs the pedagogy, never sets the syllabus, and never replaces human relationships.
2. No AI generation of Cornish
Current chatbots make too many errors. Cornish remains in human hands: teachers, speakers, communities. When AI does produce Cornish text (even unintentionally), teachers are equipped to recognise and correct errors.
3. Minimal data exposure
Only share material you are comfortable with. No personal learner data. No confidential information.
4. Transparency for learners
Learners should always know what role AI is playing (if any). For example: "I used a chatbot to help organise these activity ideas" or "This grammar explanation was drafted with AI support, then checked by me."
5. Respect for community and culture
AI does not take over creative tasks, storytelling, or cultural transmission. These belong to people.
6. Teacher autonomy
Some teachers may prefer no AI involvement at all. Our approach respects this by making AI entirely optional. Teachers who choose purely human methods are equally supported.
AI models run on large computer systems that use electricity. Every message you send to a modern AI service requires energy, more than a normal Google search, but usually far less than streaming video or long-distance travel.
Two key principles affect energy use:
Task type matters: Text-based tasks use far less energy than generating images, audio, or video
Length matters: Long chats use more processing than short, focused queries
To put this in perspective, here are illustrative comparisons based on current published research. These figures are approximate and vary by model and provider.
A single short text-based AI query (a few sentences):
≈ 2–3 Google searches, or 10–20 seconds of video streaming, or 1–2 minutes of LED lightbulb use
A long, multi-turn text conversation:
≈ watching a 5–10 minute HD video on YouTube
Generating a high-resolution AI image:
≈ roughly similar to streaming a 20–30 minute TV episode (significantly more energy-intensive than text)
Occasional, purposeful text-based use (our recommendation):
≈ the same weekly energy as sending a few emails or browsing a few webpages
One hour of continuous AI usage:
≈ boiling a kettle 2–3 times
Key point: The text-based, purposeful AI use we recommend has an environmental impact similar to other small digital activities. Image generation and heavy, continuous use scale up quickly, which is why we focus on short text queries for specific pedagogical tasks.
It's also worth noting that the energy cost of NOT having AI support, such as teachers driving to multiple in-person training sessions or printing extensive paper materials, can be considerable.
We (Change Laboratory group) encourage low-energy practices for three reasons:
1. Environmental responsibility
Short text prompts and single-task usage keep energy use similar to everyday digital habits, not high-intensity computing.
2. Pedagogical integrity
AI should never replace teacher expertise or human connection. Limiting AI helps keep lessons focused on real communication, not digital shortcuts.
3. Transparency and control
Teachers know exactly what AI is being used for: small tasks, prep support, improving organization — nothing hidden, nothing automated behind their back.
Short, one-off tasks, not long exchanges
Text-based queries only: no image, audio, or video generation
English-only suggestions, never Cornish-language content
Teacher prompts focused on pedagogy, not heavy content creation
Active encouragement to build a shared prompt library so teachers reuse effective prompts rather than each generating new outputs
Promotion of CLT practice, meaning most classroom time is human speech, not screens
Regular reflection on whether AI is still serving pedagogical goals or becoming habit
Our aim is to help teachers save considerable time on preparation , not to create a long-term dependence on high-energy tools.