🧠 Cognitive Offloading with AI

[Authors note: I’m going to be publishing a bit less frequently than usual for a while, as my doctoral studies are ramping up. Luckily my dissertation focuses on the use of AI (especially chatbots and conversational agents) in education, so still plenty of fodder for this newsletter.]

What happens to education and the way we do our jobs—what happens to us–when we turn over our memories to machines? Trying to reduce “cognitive load” (or the sheer amount of stuff we have stored in our brains and need to use for a given task) has been a problem learning designers and learning scientists, app and UX designers, and other practitioners such as teachers have wrestled with for decades. But could we circumvent it all together by handing off our “memories” to an app or chatbot? 

That might sound a little far-fetched, but it is already happening and on the horizon. With large language models (like ChatGPT) more and more capable of retaining ‘memory’ from previous conversations, what will that mean for how much information people need to memorize or hold in our heads? And with so much of educational testing predicated on short term memorization, retrieval, and regurgitation, should companies like Pearson, ETS and other test prep apps be quaking in their scantron boots? Some would argue that if students can use a chatbot to help them remember ‘temporary’ things, it is no different than them using a calculator to help them with quick calculations. Right…?

[It makes me wonder if this type of “cognitive offloading” could even lead to our brain structure changing one day (centuries or more from now) to shrink memory centers like the hippocampus and enlarge other regions–like language processing in the frontal and temporal lobes? That’s a bit out there in terms of the time horizon or my next sci-fi novel, but “cognitive offloading” or letting apps “remember” things for us is already here and controversial, to be sure, and we aren’t sure where it will lead. One professor from the University of Monterrey believes handing off our memories via cognitive offloading to AI could “turn us into imbeciles” and represents a “catastrophe” in the making for society. He and some fellow researchers liken it to our muscles losing their ability to function if we don’t exercise them. Yikes!]

But back to less dystopian possibilities! The “forgetting curve” is a well known phenomenon. As is the short term memory number limit–a simple example is the fact that we have 7-digit phone numbers because that is an amount of digits or numbers that most humans can hold in our short term memory, or encode into longer-term memory as necessary for rapid recall (RIP my Granny whose house phone # from the 1990’s I can still remember without pause). With these limits built into our human brains like a max speed on a car, does it make sense to “collaborate” with a tool to help us with memory storage and retrieval? What would this mean when the machines / servers / internet goes down for minutes or hours at a time? What if we hit our memory storage ‘caps’ for a month on a company’s freemium plan? But what if it enables people to tap into tasks and projects they’d never have been able to access on their own? Could this be a boon for equity?

How might you, in your personal life or profession, utilize a “memory keeper”? Or do you plan to be in the camp that decides to keep something as personal as memories out of the cold “hands” of a machine? For those who are in the “no” camp… might you not have already offloaded some of your cognitive load to Evernote, Google Keep, or Microsoft OneDrive over the years to hold information or a memory for you until you need it later? I’ve definitely saved pictures of favorite wine bottles I’ve discovered, for instance, in a digital notes app that I can pull up whenever wine shopping. Is this all that different? In what ways?

Practical Applications:

• For educators and trainers: reflect on the tasks and assessments you create or administer. What percentage of these call for basic recall of factual information (aka: a single right answer, a single date or historical figure, etc). If students were to utilize a memory keeper in your class or program, how might that change the way you construct your activities or projects?

• For career counselors and workforce managers: think about careers that focus heavily on short-term memory or number retrieval, such as air traffic controllers or waiters–who have to keep copious amounts of short term prices, orders, or flight patterns in their heads at every given moment. What might these careers look like if cognitive offloading becomes more popular? How might the candidate profile and skills requirements shift and what would training for these programs look like aided by a cognitive offloading tool or app?

Who Else is Talking About This:

• Morrison and Richmond (2020) studied cognitive offloading and whether or not people with greater short term memory recall or those with poorer short term memory recall could benefit more from cognitive offloading (eg to phone apps). They found, counter to some other research, that: “cognitive offloading may be a valid compensatory strategy to improve performance of memory-based tasks for individuals with a wide range of memory ability”. Meaning all of us.

• Tom Gruber (co-creator of Apple’s Siri) thinks cognitive offloading may be one of the best uses for AI (Wired). “The basic pieces of cognition, the fundamental one is memory. It's the functional inner loop, the basis of almost all of our inference. Almost all of our daily cognition or computation is memory-based. We can do so much augmentation with just memory. It's just one of the things that humans aren't so hot at.”

Learning in the News:

• ChatGPT Edu = a version for College Students, Faculty, & Researchers (OpenAI)

  • Comes with text, vision, and data analysis features powered by ChatGPT-4o

  • Built on the success of AI in higher ed at ASU, Penn, Columbia, etc.

  • Offers personalized tutoring, resume review, grading support for profs, grant writing support for faculty and researchers, and more

• AI / chatbots can now detect sarcasm, which could help us engage more naturally with AI assistants for a range of tasks (The Guardian)

• More teachers (and their unions) are ringing the alarm: AI should not replace teachers! (EdSurge)

• A taxonomy for AI use in education for students (eSchool News)

  • Lays out full ban, partial ban, no decision, partial use, full use scenarios 

Workforce Roundup:

• Google, Meta, Microsoft, Oracle, and other companies: we have to watch out for misinformation and harmful content with AI (Business Insider)

• Graduates concerned AI will impact their job prospects (Staffing Industry Analysts)

• AI could raise global GDP by 7 trillion dollars over the next 10 years. (Forbes)

  • That will only come to pass if we train and harness a workforce ready to work with AI

  • Most business leaders feel there is a moderate to significant skills gap at their companies, according to Deloitte. 

  • Only 14% of front-line workers have received any training in AI as of early 2024.

  • Companies have to step up to the plate with dedicated training and support for professional development. 

Term of the Week:

💡 Attention Mechanism: Attention mechanisms help neural networks focus on the most important parts of the user’s question or data. This improves their performance on tasks like language translation and image captioning. Example: Translating a sentence by focusing mostly on the relevant words in the input by giving each piece of the question a higher or lower weight (by estimated user importance).

If you are enjoying AI Apprentice and want to see it stay free forever, please share it with your networks! Link: AIapprentice.info/subscribe