The internet forever changed how educators and students access and share information. Now, ubiquitous access to natural language-based artificial intelligence (AI) tools will create previously inconceivable opportunities for educators and students to synthesize that information to generate ideas and design solutions. The trajectory of this innovation is easy to summarize but difficult to wrap your head around: The capacity of AI to do more than we can imagine will advance at a rate more quickly than we can understand. What can leaders possibly do to help educators effectively navigate the ambiguity—as well as the sheer potential—of the changes that await?
In our book Five Levers to Improve Learning: How to Prioritize for Powerful Results in Your School (ASCD, 2014), Jim Rickabaugh and I argue that how we respond to innovation is more important than the innovation itself. Regardless of what the innovation is, there are patterns among the assumptions we tend to make, the knowledge we apply, the strategies we use, and the associated leadership behaviors that will ultimately determine whether our efforts are irrelevant or lead to improvement.
To respond to an innovation as potentially transformative as AI, leaders should ask three questions: (1) What has changed? (2) What is the magnitude of the change? (3) Given the magnitude of change, how should we respond so that we align strategy to intended results?
What Has Changed?
Artificial intelligence describes a process through which a computer can be trained to analyze data to generate a unique, logical solution, even though it hasn't been programmed to provide that specific solution. We've had access to AI on our phones and computers for years. AI is how Google "knows" the next word you'll likely type and how Alexa can "understand" what you are saying when you ask about the weather.
The breakthroughs in AI in just the last few months appear to have come out of nowhere, but they are the culmination of 80 years of work.
The breakthroughs in AI in just the last few months appear to have come out of nowhere, but they are the culmination of 80 years of work by mathematicians, linguists, and computer scientists. A simple way to think about what has changed given recent advancements in AI is to compare a familiar tool to the new innovation. For example, let's compare a search engine to ChatGPT.
A search engine is a gateway to the internet. Search features are built on a layer of computer code that sifts through structured data, such as web pages denoted by keywords, and lists them. At its most basic, a search function uses a "look-up" and "fetch" approach to efficiently access content that has already been created and labeled. If the web page includes specific key words, then it will be presented in the list. In recent years, search engines have been trained using machine learning, a subset of AI that applies code and algorithms, to sort and filter data even more efficiently.
In a familiar classroom scenario, 5th grade students learning about recycling could use a search engine to find articles about recycling. They might enter "importance of recycling" into the search bar, and they would get links to millions of web pages with information written by people on that topic.
ChatGPT, by contrast, uses deep learning, a subset of AI that uses transformers (think interconnected hierarchies of code, not cars that can turn into robots) to detect patterns across thousands of layers of if-then scenarios. Large language models (LLMs) are a specific type of deep learning model trained to recognize and predict patterns in language. The most powerful LLMs have been pre-trained with so much data that they can interact with and generate text for nearly any purpose and in any format. ChatGPT is an example of a generative, pre-trained transformer. The most important thing to understand about LLMs is they aren't search engines. They seek patterns and predict the most likely sequence of words to generate a unique, nuanced response.
Applying this technology to our classroom scenario, the 5th grade students writing about recycling could prompt ChatGPT4 (or another LLM tool) to "Write a short, persuasive essay that provides three reasons to use fewer disposable plastic products using the words renewable, nonrenewable, biodegradable, and organic. Write this at a 5th grade reading level, by a student who loves swimming." The program will write an essay that meets those exact specifications.
This means students can now use ordinary language to direct technology tools to generate, summarize, synthesize, and create unique, sophisticated responses or products in a manner that previously could only be accomplished by humans who had learned that content knowledge or developed specialized skills. This is an unprecedented change in the history of information science—and education.
What Is the Magnitude of Change?
Magnitude of change is a term used in leadership theory to describe the extent that a change requires individuals to use different knowledge or skills, or adopt new ways of thinking, to achieve a goal. Three ways to describe different magnitudes of change are:
- Status quo management—You apply your current knowledge and strategies to available resources or innovations to maintain results.
- Transactional change—You apply your existing knowledge and strategies, but more frequently, or to different tools, to improve results and efficiency.
- Transformational change—You apply new ways of thinking and different strategies in using new tools to maintain or improve results and create new pathways or products.
These three terms do not represent a hierarchy of bad to better magnitudes of change. Status quo management is essential in many scenarios; it ensures the buses arrive on time and the lights work. What is most important for leaders to understand is that navigating different magnitudes of change requires different leadership behaviors. The risks of misalignment between the magnitude of the change and a leader's response to that change often determines whether initiatives will succeed or fail. For example, trying to preserve the status quo during a period of transformational change can leave teachers and students preparing for a world that no longer exists. Similarly, applying transactional solutions to transformational challenges typically results in increased effort without better outcomes.
Applying transactional solutions to transformational challenges typically results in increased effort without better outcomes.
New AI tools will present transformational opportunities—and challenges—for teaching and learning. By considering the distinctions among status quo, transactional, and transformational approaches to adapting to a world with ubiquitous access to sophisticated AI tools, teachers and administrators can more mindfully invest their thinking and effort into using these tools in a manner that expands and supports, rather than undermines, teaching and learning.
How Should We Respond to Align Strategy to Intended Results?
We often think of maintaining the status quo in our schools or classrooms as a passive approach that requires little effort. However, maintaining the status quo in response to a transformational change can be actively counterproductive.
Given the unknowns of ChatGPT upon its release, for example, many schools simply blocked access on their networks in an effort to maintain the status quo. The rationale for this approach may have initially been justifiable, but this is not a long-term solution for navigating a transformational challenge. For one thing, blocking local access to something that's widely available will result in a widening of the digital divide; students with access to these tools outside of school will use them anyway. Furthermore, blocking access means students will not have a chance to learn the skills and strategies necessary to use AI tools in productive and responsible ways. That's an odd way for educational institutions to treat a technology that is expected to play a significant role in students' futures.
Using a transactional approach with a transformational tool like AI can be even more problematic. When a new innovation becomes accessible, users typically apply their existing assumptions, knowledge, and strategies to the new tool to accomplish a goal more efficiently. For example, many users' first instinct is to use ChatGPT as a souped-up search engine. They type in a query on a topic and wait for the results. Not surprisingly, these users may leave the experience confused about all the hype.
A more productive, though still transactional, approach to using AI is for students and teachers to use it to generate specific ideas for teaching and learning. For example, teachers could use AI to generate lesson plan ideas based on very specific criteria, or to come up with tips for activities and projects that meet particular criteria. Similarly, students could use AI to generate ideas for projects or to create mnemonic devices to memorize academic content. What had required a web search or a visit to Pinterest, followed by some selecting and synthesizing, can now be accomplished more precisely and efficiently with AI.
But what happens when the transactional assumptions, strategies, and intended results are out of alignment with the transformational nature of the technology? Let's consider two examples.
Many educators fear students will simply use AI for cheating, essentially a transactional activity. A transactional counter-response to this very real problem is to modify plagiarism and acceptable use policies by adding language about AI and ramping up penalties for misuse. Before analyzing this use and response, let's consider a second example.
Teachers today are overwhelmed. Any opportunities to create efficiencies in tasks such as lesson design, creating assessments, and grading student work are justified and welcomed. As a transactional tool, AI will be able to accomplish each of these tasks. Students could seek similar efficiencies to do their work.
Now carry these transactional approaches of both students and teachers using AI to fulfill basic processes of schooling to their logical conclusion: teachers use AI to design assignments, students prompt AI to complete the assignment, and teachers prompt AI to grade it. What started as a quest for efficiency becomes a process of data-entry specialists tweaking a series of algorithms to improve the AI's automaticity. This could be the worst-case scenario for AI use in schools, with students and teachers relying on AI to churn out more work, but that work being soullessly devoid of relevance or meaning.
These examples reveal some truths about the complexities of helping others navigate change. When new, powerful tools allow people to reach for quick, transactional fixes, many will be satisfied in doing so. Absent the opportunity to develop new understandings about the nature of the work, a transactional application of old strategies applied to new tools is seen as the only possible next step. Ultimately, this misalignment of a transactional approach to a transformational challenge will result in a diminished capacity to teach and learn.
Transformational Approaches to AI
During times of transformational change, it can be difficult for leaders to acknowledge they don't have all the answers. Absolve yourself of this responsibility. Leading transformational change requires adopting new ways of thinking about the nature of the work, challenging long-held assumptions, and helping others break free from routines that had worked well in the past but may no longer be relevant. In this context, asking big questions is more important than providing simple answers.
Leading transformational change requires adopting new ways of thinking about the nature of the work.
It should not be surprising that a transactional question about AI such as "How can AI make lesson planning, assessing, and grading more efficient?" will yield transactional answers. Effective leaders understand that in times of transformational change, asking the right questions is more important than having all the answers. To reframe the conversation and create new possibilities, we must ask big, open-ended questions. For example: "If you could use AI to provide individualized, one-on-one teaching and tutoring to students each day, what opportunities would you provide to each learner?" Such a question could generate ideas like the following:
- Create opportunities for content to be accessible to each student: Teachers can paste a passage of text at a 10th grade reading level into an AI tool and ask it to rewrite the passage at a 3rd grade reading level, or in a different language, in a matter of seconds. The possibilities for differentiation and tailoring content to students' individual needs or circumstances are unparalleled.
- Meet learners where they are and engage them in personalized formative assessment: A student could follow a set of prompts from the teacher to have a conversation with an AI tool about an academic topic. The tool could discern what understandings and misunderstandings the learner has on the topic. Then, the AI tool could coach the learner to address specific misconceptions, provide a short passage of clarifying text, create a quiz, and provide instantaneous, formative feedback.
- Create specific exemplars and non-exemplars that could help students assess their own work: We learn complex concepts by looking at examples and non-examples. When prompted effectively with well-designed success criteria from rubrics, new AI tools are remarkably adept at providing relevant examples that can show students what to aspire to, and what to avoid, when writing or engaging in other complex tasks. Teachers can tailor these examples to draw on students' existing strengths and schema to help them focus on the transferable elements of quality.
These responses are transformational because they discard old assumptions about the inter-relationships among technology, information, teachers, students, and learning. A question that invites teachers to think differently about how AI might be used as a tool to transform their capacity to provide targeted, relevant, instruction to each student can help them see beyond the transactional, urgent busyness that can fill a school day (see Figure 1). This vision, not the technology itself, is what will ultimately determine each teacher's capacity to use these tools in transformative ways.
Positioning for Change
Leading transformational change takes time. We don't yet know how to fully harness the transformative power of AI to improve teaching and learning. And that is OK. What leaders can do is create conditions that support transformational ways of thinking about how the innovation can be utilized. They can challenge existing paradigms, engage in dialogue around big questions related to teaching and learning, create intentional opportunities for teachers to explore new tools, and acknowledge that our response to AI is what will ultimately determine whether AI improves or diminishes our capacity to teach and students' capacity to learn.
Five Levers to Improve Learning
The key to success is not doing more work and making more changes, but doing the right work and making the right changes, say Tony Frontier and James Rickabaugh.
