A common misconception about AI literacy is that it’s about skills – teaching students to use AI to run faster on existing academic trajectories, to achieve professional goals, or to perform at higher levels. While it certainly is important to know how to use AI in academic settings to accomplish specific tasks and assignments, this is merely scratching the surface of AI literacy.
Limiting your school’s definition of AI literacy to this narrow scope misses the critical idea behind high-quality AI literacy programmes and curricula: to teach students how to identify and cultivate their own voice (and by extension, their own identity) in a world that is increasingly saturated with the sameness of AI-generated content.
If you haven’t already been doing so, now is the time to start keeping your finger on the pulse of your students’ voice and expressions. Why? Because credible evidence is starting to emerge that demonstrates that words most commonly used by AI chatbots have started to infiltrate human language – a new slippery slope for any educational institution interested in nurturing holistic human beings.
A recent study found that AI-assisted writing (i.e. LLM-assisted writing) has penetrated multiple sectors of society at significant levels, beyond the K–12 sector. Approximately 18% of financial consumer complaints, up to 24% of corporate press releases, nearly 14% of United Nations press releases, and up to 15% of job postings showed significant use of AI assistance, mirroring similar widespread adoption that previous studies had found in academic research globally.
Other studies on LLMs and writing have found that LLMs tend to choose biased language that frames certain genders or people in skewed manner based on biased training data, including being predisposed to generate stories about gender normative topics (family for women, politics for men) when prompted to write a story about a male or female character; or associating certain words and attributes (like “friendly” or “weak”) with certain descriptions of characters’ appearances or demographic groups.
WHAT DOES IT MEAN?
First, there is the slow creep towards homogenisation: the risk that AI-assisted writing will, slowly but surely, make our writing more similar, and make our individual voices less unique, and perhaps even more biased. As LLM-inspired language continues to proliferate in the vicinity of our young writers, it is important to encourage them to choose their own words for what they are describing or depicting, with intention and thought, so that they do not default to perspectives provided by training data that scrapes every inch of the internet.
Second, there is the inevitable undertow of homogenisation: a potential loss of meaning and purpose. As the range of written expression narrows, so can the emotional complexity and experience of the individuals writing the words. When we lose words, we lose nuance. And once we lose our voice, our individuality and sense of purpose begins to erode, slowly but surely.
As a school leader, it is time to consider which kind of AI literacy programme your institution is investing in: one that teaches students to use AI as a shortcut and encourages mimicry to fulfil mid-term academic goals, or one that guides students to interrogate and affirm their own sense of identity as learners, using AI-augmented learning as a jumping-off point for discussions of lifelong learning and purpose? To explore further how AI-generated text reflects biases in the literary canon, see ‘Written in the Style of’: ChatGPT and the Literary Canon.
