Let’s not lose our ability to deeply process information
Our mind needs to keep its muscle strong for critical thinking
Image by cookie_studio on Freepik
Have you ever found yourself having read a news article only to realize you can't recall much of its content afterward? Or perhaps you've attended a lecture and struggled to remember what the speaker said? Maybe you've been in a meeting and, despite your best efforts, can't remember many details or the decisions made?
You might initially think you simply weren't paying attention. But if we look a bit more closely and examine what’s going on, it might well be that though you were physically present at the meeting you just weren’t processing the information very deeply.
Fifty-two years ago, psychologists Fergus Craik and Robert Lockhart at the University of Toronto introduced the concept of “depth of processing.” They suggested that we perceive information in different stages. Initially, we focus on the sensory or physical characteristics of what we encounter. For example, when listening to a speech, we might notice the pitch of the speaker's voice or if they stumble over words. However, if we focus solely on these superficial aspects, we may not fully grasp the content of the speech.
On the other hand, we can engage in deeper processing by connecting the ideas presented in the speech with our existing knowledge or by asking ourselves questions about the ideas. This deeper level of engagement allows us to reflect on what was said and compare it with our previous experiences.
A typical experiment to test depth of processing, which many psychologists have done over decades, is as follows. Subjects are brought into a laboratory and given a list of words. In one condition, they are asked to focus on structural aspects of the words, like being asked if the word ‘baboon’ contains the letter “a”. In another condition, subjects are asked to think about the words semantically: is the word ‘baboon’ a living thing? In the first condition, subjects are just focusing on physical aspects of the word (does it contain a certain letter). In the latter condition, subjects are thinking about the meaning of the word, enabling them to connect it semantically to their other knowledge. The results consistently show that memory performance is superior when participants engage in semantic processing, where they think about the meaning of the words, compared to merely focusing on their physical or structural features.
Neuroscience studies support these findings. When subjects engage in deeper information processing, PET scans and fMRI images show regions in the brain associated with more brain activity compared to when subjects engage in shallow processing.
Craik and Lockhart argued that there really is little advantage in storing things in our memory that are processed in a shallow, preliminary way, such as the pitch of a person’s voice, unless it stands out as particularly unusual. But when we deeply process information, we ask questions about it, make connections with existing knowledge, and generate new ideas, creating a stronger memory for it.
When we encounter things that we’re familiar with, then we process them in a deeper way faster. That’s why if you happen to know Spanish, you might find it relatively fast to learn some Italian words especially if they are similar to words that you already know in Spanish.
Craik and Lockhart suggest that levels of processing occur in a continuum. In other words, there’s no abrupt cutoff between what is preliminary shallow processing and what is deep processing. The more deeply something is processed, the more enduring its memory trace becomes. So if you’re preparing for an exam, then you need to process the material more deeply to remember it. You shouldn’t just read it, but should also ask questions about it, create diagrams of it—in other words, to really work with the information.
The findings from depth of processing studies have broad implications for understanding how we handle information in daily life, especially in our digital age. We're constantly bombarded with information and at the same time we have tools at-hand that can process that information for us. It sounds like it should work well, right? Maybe not. I worry that we are heading down a road where we are losing opportunities to process information in a deeper sense.
The prevalence of Large Language Models (LLMs) present us with a challenge. On the one hand, such tools ease our cognitive load--the machine is processing the information, not us humans. Let’s say that a student is given an assignment to write a report on the effects of climate change over the last 10 years. The student types in a prompt to the LLM and asks it to write the report. Hopefully the student reads the output to correct hallucinations before turning it in .
But contrast this with the scenario of the student reading actual articles, analyzing graphs of trends, and formulating their own interpretations of the topic. They are engaging in deep processing. Can you guess in which scenario the student will have better comprehension and memory retention? In just quickly scanning the LLM output without doing the actual reading of the raw material that went into the report, the student has engaged in a much lower level of processing along that continuum described above.
LLMs save us time by summarizing and writing information for us. But while we may gain time, we are losing out in other ways. We are losing the opportunity to process information in more depth which makes us understand the material better and to retain it longer. You might say that this is fine because then we have extra time to do things that count. But will this really happen? In a previous Substack post I wrote about the slothful mind. In other words, the human mind is inherently lazy and we tend to take the path of least resistance.
Our tendency to opt for the path of least resistance can further exacerbate this reliance on AI. Who wants to write a monthly report which is busy work? If we relegate that to AI, then what a boon for us. What I worry about though is that while we may have the best intentions to use LLMs in a limited way, it’s so easy for us to slide down that slippery slope. We may rely on using AI more and more to process that information for us, because we humans want to do what’s easiest. As a result, we risk losing the ability to reason critically, because we’re not doing it-- AI is. Deep thinking is a skill that we cannot afford to lose.
We are doing ourselves a disservice when we blindly use AI to process information. We need to carefully consider which information we can relegate to AI and what we can best benefit from by engaging directly with the information. We risk weakening our mind’s muscle for critical thinking if we outsource much or all of our work to AI to do the processing. The only way to keep our mind’s muscle strong is to regularly practice deep processing of information. AI is a tool designed to help us, but let’s use it carefully. Let’s preserve our mind’s powerful muscle so that we can be present, say at a meeting or a lecture, so that we’re not there just physically but we’re also there with full mental awareness.
You can learn more about depth of processing and our attention in the digital age in my book Attention Span.