AI and the Authenticity Problem
AI is forcing us to question what's authentic. That might be a good thing.
Merriam-Webster’s word of the year is “authentic.”
In their citation, they write that because of the emergence of AI, “the line between ‘real’ and ‘fake’ has become increasingly blurred.” They add that authenticity is about much more than real or fake; it also means “true to one’s own personality, spirit, or character.” They identify a troubling problem with authenticity in the modern world: with “‘authentic content creators’ now recognized as the gold standard for building trust, ‘authenticity’ has become a performance.”
We live in a world where both humans and AI are performing authenticity, making its true version harder to find and more important to prioritize and celebrate. In education, this conversation ranges from the practical problem of knowing when a student’s work is truly their own to the philosophical problem of how to preserve authenticity in education when it’s so much easier to just perform it.
For the first time in a few years, I revisited one of my favorite pieces on authentic education, Grant Wiggins’ 1989 essay “The Futility of Trying to Teach Everything of Importance.” It’s kind of astonishing to read it now. Almost 35 years after its publication, it seems more relevant.
This is the key part for me:
“The inescapable dilemma at the heart of curriculum and instruction must, once and for all, be made clear: either teaching everything of importance reduces it to trivial, forgettable verbalisms or lists; or schooling is a necessarily inadequate apprenticeship, where ‘preparation’ means something quite humble: learning to know and do a few important things well and leaving out much of importance. The negotiation of the dilemma hinges on enabling students to learn about their ignorance, to gain control over the resources available for making modest dents in it, and to take pleasure in learning so that the quest is Lifelong.
An authentic education will therefore consist of developing the habits of mind and high standards of craftsmanship necessary in the face of one's (inevitable) ignorance.”
In what Wiggins calls a “modern curriculum,” we let go of the idea that there is a fixed set or amount of knowledge students must have and instead prioritize teaching students a process of knowledge acquisition and application that empowers them not just to know things, but to understand what they don’t know, what they want to know, and how to learn about it. This inquiry-based approach, which I’ve written about before, launches students into an ongoing exploration of deeper and harder questions driven by their interests and increasing knowledge.
For Wiggins, the unit of an authentic education is the question, not the fact, and some of those questions must come from students because, ultimately, the outcome we seek is student-driven learning. The implications for curriculum and assessment are significant: we abandon content coverage for deeper exploration of fewer ideas, and responsiveness and flexibility in curriculum replace scope and sequence.
This means we will work with content differently. It’s fascinating to read this passage with AI in mind:
“The teacher must have access to material that offers a variety of specific inquiries to pursue, with suggestions on how to deepen student responses… The textbook, instead of being the syllabus outline and content, would be a reference book for student and teacher questions as they naturally arise. Like the music or athletic coach and the vocational education teacher, the classroom teacher's job is to help the student ‘play the game’ of the expert, using content knowledge, as contextually appropriate, to recognize, pose, and solve authentic knowledge problems.”
In AI, we have at our fingertips a reference tool exponentially more powerful than the textbook that can not just help students and teachers locate knowledge when they need it, but can also assist them in identifying and fleshing out the “authentic knowledge problems” that define expertise in a given field.
Of course, AI could be used to short circuit this process by presenting work that appears authentic without any actual learning behind it. It can also, as has been well documented, be wrong. Its power is double-edged. I do not see this as a compelling reason to avoid using AI, nor do I believe that it’s possible to prevent all students from using AI to cheat. I do believe that we can design assessments that achieve Wiggins’ goal of sparking the desire to learn while deepening students’ understanding of what it means to develop expertise both with and without AI.
Can AI enable authentic assessment?
Wiggins offers four design elements for an authentic education: “(1) equip students with the ability to further their superficial knowledge through careful questioning, (2) enable them to turn those questions into warranted, systematic knowledge, (3) develop in students high standards of craftsmanship in their work irrespective of how much or how little they ‘know,’ and (4) engage students so thoroughly in important questions that they learn to take pleasure in seeking important knowledge.”
Consider these examples, which I think meet Wiggins’ criteria, whether they are using AI to enhance existing assessment designs or to launch students in brand new directions:
Take a few minutes to watch this video from an English teacher explaining how she is blending analog in-class work with guided use of chatbots to support students in deconstructing the writing process, creating and analyzing their own writing, and, as she shares in her follow up, discovering what “voice” is and why it matters. (Thanks to Maha Bali for sharing this and so many other examples of assessments for AI literacy.)
I was a middle school and high school English teacher for 12 years, and if I were still in the classroom, I’d spend at least two or three classes asking students to create bots in Poe (or, if we all had ChatGPT-4, creating GPTs). From a writing and communication perspective, it requires precision of language and syntax, responsive editing, critical reading, and resilience. From an AI perspective, it reveals how challenging effective prompting is and how much human thinking and intervention AI still requires to succeed. Plus, it’s a truly authentic assessment: either your bot does what you want it to, or it doesn’t. Lance Cummings has a good overview of how to create a bot in Poe.
I recently heard that students at the American School in Japan were using Poe to build Japanese-English translation bots and studying how well different chatbot platforms (GPT, Claude, Google’s PaLM) perform on translation tests. I like the idea of positioning students as researchers who study AI: it’s rigorous work that builds agency. AI tools like Consensus and Elicit and ChatPDF are designed to support research and interactions with long documents.
Maria Dikcis imagined playing the game Exquisite Corpse with multiple chatbots. You could do versions of this in many different ways with many different age groups. The reflective questions at the end are, to me, evidence of this assessment’s authenticity and value.
Fay Short’s students suggested using ChatGPT to create job descriptions, application forms, and simulated interviews to prepare for post-graduation life. Short’s assignment is #60 on this list of 101 Creative Ideas to Use AI in Education.
For folks who already have some AI skills, Leon Furze used generative artificial intelligence to build an entire world in the mode of science fiction. His detailed description of his process reveals all the pillars of authentic assessment: a novel problem without a predetermined solution, application of prior knowledge and acquisition of new knowledge, using critical thinking to navigate ambiguity and uncertainty, and an artifact of work that reflects the interests and perspective of the learner. Plus, it’s fun.
In the mode of science nonfiction, a physics teacher recently shared with me an astronomy unit he did with students using ChatGPT to create apps that plot data so that students could pursue questions about habitability of exoplanets, luminosity of stars, and more. The potential for AI to make sophisticated data analysis and programming more accessible to students in a variety of disciplines, not just computer science, is exciting to me.
Are we asking the right questions about AI and authenticity?
I don’t share these examples to suggest that everyone now needs to be integrating AI into every assignment. Rather, I think they illustrate that AI does not determine the authenticity of an assessment. Humans do. If we are worried that AI makes an assessment less authentic, then we should not only be worried about AI. We should also be worried about the assessment.
Towards the end of his essay, Wiggins writes that courses at Central Park East High School in East Harlem, NY, USA, are designed around five questions of authenticity:
Whose voice am I hearing? From where is the statement or image coming? What's the point of view?
What is the evidence? How do we or they know? How credible is the evidence?
How do things fit together? What else do I know that fits with this?
What if? Could it have been otherwise? Are there alternatives?
What difference does it make? Who cares? Why should I care?
Instead of asking, “How AI-resistant is this assignment?” what if we asked these five questions instead?
Links!
Bryan Alexander has curated an enormous, crowdsourced list of resources on thinking critically about AI.
This multimedia piece, “How AI Reduces the World to Stereotypes,” would be an excellent spark for conversations and reflections about AI’s ongoing issues with bias.
A crowdsourced think tank of AI research and projects from across industries that are aligned to the United Nations Sustainable Development Goals.
If you haven’t yet experienced Angie Wang’s lovely graphic essay about AI, “Is My Toddler a Stochastic Parrot?”, you should. Put it on the biggest screen you can.
Sara Elaine Eaton on how AI is moving us into a post-plagiarism world.
Marc Watkins on why AI literacy should be the lodestar of AI work in schools (included are many links to how AI is expanding beyond chatbots).
Online gradebooks are a case study for schools on the importance of intentionality, ongoing education, and responsive policy when deploying new technology: “Snowplow Parents Are Ruining Online Grading.”
Learning on Purpose is taking a break for the rest of 2023 while I work on other projects and spend time with family and friends. I hope you’ll use and share the archive to explore older posts.
Also, let me take this opportunity to say: thank you. When I started this four months ago, I was just looking for a place to keep myself thinking and writing as I went through a professional transition. Since then, more than 800 people from 50 countries have subscribed and Learning on Purpose posts have been viewed more than 20,000 times (one was even in The Marshall Memo!). Thank you for reading, for subscribing, for sharing, and for connecting with me. I’ve learned a ton.
If you want to learn more about my consulting work, offer feedback, or just talk shop, you can reach me at eric@erichudson.co or connect with me on LinkedIn. As always, it’s free to subscribe to Learning on Purpose. I hope you’ll subscribe and share. See you back here in January. Happy New Year!