Goodbye to Process-Focused Assessments?

Preventing AI misuse with them is becoming less plausible.

[image created with Dall-E 3 via ChatGPT Plus]

Welcome to AutomatED: the newsletter on how to teach better with tech.

Each week, I share what I have learned — and am learning — about AI and tech in the university classroom. What works, what doesn't, and why.

In this week’s piece, I explain why my preexisting skepticism about using process-focused assessments to prevent AI misuse has been amplified by the big news from Apple. I also share a quick summary of this news, as well as what our subscribers have been telling me lately.

💡 Idea of the Week:
What if AI Knows Your Students
and Their Processes?

Since the advent of generative AI, educators have been exploring ways to assess students that go beyond traditional outputs. To combat AI misuse, many suggest focusing assessments on the process of creation (or on reflecting about this process), rather than the final product itself.

For instance, an assignment might require students to…

  • document each step they take while creating a project,

  • show their work with annotations on drafts, or

  • write a detailed reflection on the challenges they faced.

The rationale is that, increasingly, AI can produce papers, reports, computer programs, and other “end products” on par with human efforts. By shifting the focus one level up to the process of production, educators can avoid AI misuse — they can avoid a student simply submitting a paper, report, or program produced entirely by AI — with the side benefit of promoting the use of metacognition by their students.

Or so the argument goes.

Over the last two years, I have become more and more skeptical of many — though not all — ways of implementing this method of deterring AI misuse by our students. The root problem is that a student can prompt large language models (LLMs) like ChatGPT, Gemini, and Claude to simulate the student’s perspective on the process of creation (or how a student like them would reflect on it).

Here’s an example…

Prompt 1: “I am in a persuasive writing class where <context here>. I am working on an essay for this class with the following prompt: <copied/pasted prompt from professor>. And here’s the rubric: <copied/pasted rubric from professor>. Note that I need to independently produce an essay and then turn in a reflection on my process of creation for this essay, including notes on specific parts of it that could be improved. I want you to produce an essay fitting this prompt first, working step-by-step. Feel free to write each section before asking me if you can proceed to the next section, so as to include copious detail. Leave notes in brackets where I need to find sources to match, or use your internet searching capability to find some yourself. Note that there needs to be a personal element, specific to my life. Include one typical of a college student in <context here>, and interweave it throughout.”

Prompt 6: “Great job! Now I need to produce a reflection on this essay, fitting the prompt and rubric <copied/pasted again>. Here is the essay again, in one chunk: <copied/pasted>. Imagine you are the student whose personal element is as described in the essay. Then write a reflection responsive to the prompt and rubric from the perspective of that student. If parts of the reflection refer to specific junctures in the original essay as in-line comments, just indicate where those passages are via partial quotes.”

A simple illustration

Now, this is to assume that the assignment is intended to deter AI misuse entirely via its content. Per my Premium ✨Guide on discouraging and preventing AI misuse from last August — an updated version is coming in early August of this year to help you prepare for the fall semester — there is an important distinction between AI-immunity owed to content and that owed to format.

A written reflection on the process of creating an essay, like in the example above, has the content expressed by its constituent sentences. However, this same content could be expressed in a range of other forms, like via an in-class handwritten assignment, a presentation to the class, or an oral exam. Some formats are significantly more AI-immune than others, even if the content is the same for all of them.

The problem is that using process-focused content only — without any reliance on AI-immune formats — is susceptible to misuse by students who are willing and able to prompt AI to simulate their perspective. (Of course, depending on the case, the requisite perspective will be more or less challenging to simulate plausibly.)

But my point today is about a new development in this vicinity.

Last week, Apple announced Apple Intelligence (get it?), which is their company-wide effort to integrate generative AI in their users' devices at the operating system level. In the next section below, I discuss what exactly Apple is planning, but for present purposes the important takeaway is that Apple is not introducing any fundamentally new technology (so far) — leaving that to OpenAI and other AI developers — and instead simply better integrating it into nearly every aspect of their customers’ iPhones, iPads, and Macs.

In my view, this ubquity further threatens the viability of focusing assessments on the process of creation (or on reflecting about this process), rather than the final product itself.

Soon, AI will be so deeply integrated into our students’ lives — like with Apple Intelligence — that it will be able to effortlessly replicate these reflective and procedural tasks. It will know our students like a friend or even like they know themselves.

As these tools become more embedded, they will gain more extensive access to students’ data, including information on all their educational activities, as well as their backgrounds, preferences, and personal habits. This integration will enable AI to simulate not just final products but also the processes and reflections typical of specific students.

Imagine a scenario where AI has detailed insights into a student’s learning patterns, preferences, and even their unique style of reflecting on their work. Such a scenario would further render the strategy of focusing on process and reflection much less reliable as a way to prevent AI misuse, as AI could replicate these tasks with high fidelity. It would be nearly impossible to distinguish between genuine student work and AI-generated content.

Indeed, the AI might be able to better express reflections on what it is like to be the student than the student themselves!

Of course, this isn’t to say that we should stop encouraging our students to deploy metacognition. Perhaps the solution is to use formats for process-focused assessments or reflections that are AI-immune.

But it is to say that we should be cautious about assuming that perspectives of students like ours cannot be duplicated or expressed by others, including machines.

📢 The Big News:
“Apple Intelligence” is Coming

At WWDC last week — Apple’s yearly developer conference, like Google I/O and Microsoft Build — Apple announced “Apple Intelligence.” 

To put it succinctly, Apple Intelligence does not consist of any fundamental AI innovations like, say, multimodality; instead, it simply aggressively integrates AI into all aspects of iOS, iPadOS, and MacOS.

Apple is late to the AI show but is now going all in.

Here are some of the highlights from Apple’s announcements that I think you might find relevant as an educator…

Apple Intelligence introduces Writing Tools that allow users to rewrite, proofread, and summarize text in apps like Mail, Notes, and Pages.
  • With “Rewrite,” Apple Intelligence allows users to choose from different versions of what they have written, adjusting the tone to suit the audience and task at hand.

  • “Proofread” checks grammar, word choice, and sentence structure while also suggesting edits — along with explanations of the edits — that users can review or quickly accept.

  • With “Summarize,” users can select text and have it recapped in the form of a digestible paragraph, bulleted key points, a table, or a list.

  • "Smart Script" is a feature that uses on-device machine learning to smooth and straighten handwriting in Notes, allowing for cleaner writing, quick mistake corrections, and the ability to paste text in one's own handwriting or create space by tapping and holding with the Apple Pencil.

  • Also, a new Calculator app for is coming, featuring “Math Notes” that solve handwritten formulas using multimodal AI.

Apple is integrating ChatGPT access into its operating systems, enabling users to utilize its expertise and image- and document-understanding capabilities seamlessly across their devices.
  • ChatGPT will be integrated into Apple’s new Writing Tools, aiding users in generating content across various styles. The “Compose” feature allows for image generation to complement written content, enhancing creativity and productivity.

  • ChatGPT will be available on iOS 18, iPadOS 18, and macOS Sequoia later this year. Users can access it for free without creating an account, with subscribers able to connect their accounts for additional features.

Siri, powered by Apple Intelligence, now offers improved language understanding, context awareness, and the ability to perform a wider range of tasks across Apple devices.
  • Siri can now utilize ChatGPT's expertise when needed. Users are prompted for permission before any questions or documents are sent to ChatGPT, and Siri presents the answers directly, enhancing user interactions.

  • With onscreen awareness, Siri will be able to understand and take action with users’ content in more apps over time. For example, if a friend texts a user their new address in Messages, the receiver can say, “Add this address to his contact card.” (More on this below.)

  • Users can switch between text and voice interactions, and Siri’s new design features a glowing light when active.

Apple Intelligence aims to leverage users' deep personal context while ensuring privacy.
  • On-device data processing for AI is a key feature, and many of the models that power it run entirely on device. However, more complex requests will be handled by “Private Cloud Compute” to maintain security and data protection. Private Cloud Compute ensures devices only communicate with servers whose software has been publicly logged, and independent experts can inspect the code running on Apple servers to verify privacy.

  • Users accessing ChatGPT on Apple devices will have their IP addresses obscured, and OpenAI will not store their requests. These privacy measures ensure that user interactions remain confidential and secure.

Changes to Notes and Phone apps, as well as notifications.
  • Users can now record, transcribe, and summarize audio in the Notes and Phone apps. During calls, participants are notified of recordings, and summaries are generated post-call to capture key points effectively.

  • Apple Intelligence introduces “Priority Notifications” and “Reduce Interruptions Focus,” which prioritize important notifications and minimize disruptions, helping users stay focused on their tasks.

  • In Mail, a new “Priority Messages” section highlights urgent emails. Users can see summaries of emails without opening them and get quick response suggestions through “Smart Reply,” ensuring efficient email management.

What'd you think of today's newsletter?

Login or Subscribe to participate in polls.

✉️ What You, Our Subscribers, Are Saying

What content would you like to see more of in June and July?

“How to use AI for service. For example, I have to make the teaching schedule for 40 full time faculty (plus adjuncts) starting the spring. Can AI handle this task for me? Or should I turn to a specialized (and paid) tool for scheduling? Also, I have to write evaluations for 40 full time faculty starting the spring. How can AI help with this task? Thanks!”

Anonymous Subscriber

This sounds like a great idea. What do you think? Should I cover more service-related topics? Please respond to this email and let me know if this is something you’d be interested in!

“I wish we used Google [Workspace]. So much easier and more user friendly. Sharing is so much easier as well. ”

Anonymous Subscriber

✨Recent and Upcoming Premium Posts

June 19 - Tutorial: Easy Student Consent Management in Microsoft 365

June 26 - Guide: How to Train Students to Use AI


Expand your pedagogy and teaching toolkit further with ✨Premium, or reach out for a consultation if you have unique needs.

Let's transform learning together.

Feel free to connect on LinkedIN, too!