Literacy is an amazing power. But it comes at a terrible price. And no, I don't just mean memory.
Writing is Magic
Through the magic of psychometric tracery we are able to share the thoughts of fellow literates across great distances of time and space, just by reading their inscriptions. Moreover, psychometric tracery has a permanence that memory does not, so we can preserve our own thoughts more completely and precisely, for longer, by writing them down, than by remembering them. The modern bureaucratic state and firm owe their existence to writing - the world would collapse without it. This has probably been true ever since the first great cities learned the Art.
But Great Magic Comes at a Great Price
Just like meetings summon a very knowledgeable demon at the price of the temporary suspension of their participants' minds, writing comes at a price as well. The most common criticism is that literate people have worse memories. As usual, Plato said it best. I'm just going to paraphrase, if you want the original, I highly recommend reading the Phaedrus.
In Phaedrus, Plato has Socrates tell a story about the invention of writing. He says that Theuth, the god-inventor, presented his inventions to the god-king Thamus, and among them was writing, which Theuth praised as an aid to both wisdom and memory. Thamus replied that Theuth was too optimistic; writing was a drug that counterfeited memory, and actively harmed wisdom. People would be able to "recite" many true opinions that they just looked up, but out of prolonged reliance on reference texts, would have less of the understanding that would have enabled them to generate these opinions in the first place.
Elsewhere in Phaedrus, Socrates says that the true practice of philosophy cannot be written down, because to teach philosophy you cannot speak in the same way to everyone. Philosophy is not a set of opinions, it is more like a fire burning in the soul of a person, which can only be transmitted by prolonged contact in which the other person's soul can catch fire. Plato says much the same, writing in his own voice, in the Seventh Letter, which I recommend less but has the virtue of being short.
Of course, everyone ignores this and goes on to assume Plato put his philosophy into writing. Well, almost everyone.
Was Plato Right?
I don't actually think the degradation of memory is a problem. If anything, it's freed up mental space for better things to remember. Instead of memorizing facts, we can keep track of a large number of ways to obtain facts. We've increased our total power to obtain true opinions.
The understanding thing is a little more problematic.
Talking to Yourself
When you think something through in your own mind, you have access to all your own thoughts. You know what you mean by all the words you use. You can communicate with yourself in any mode - visual, auditory, tactile, nonverbal.
Verbal conversation with another person is necessarily lower bandwidth - meaning that less information is communicated at a time. In exchange, you get two separate minds, with different strengths, processing the information simultaneously. A clarifying question from your interlocutor can help you notice that actually, no, you don't quite understand what you mean by that word, or the nonverbal assumptions you were making aren't ones you endorse, or the big fuzzy thing you were confused about seems clearer when you break it down into pieces small enough to talk about.
Another problem with verbal communication is error. Disagreements about definitions or word usage often derail substantive conversations. This can be (but rarely is) addressed by frequent stopping or interrupting at the first moment someone uses a term that seems unclear. The underlying disposition of curiosity that makes this possible, and the readiness to
Latency and Throughput
Verbal communication at all has serious problems, and writing has even more. A big one is latency.
I am borrowing the concept of latency and throughput from computing. They are two measurements of how fast you can transfer information.
Throughput measures the overall rate of information transfer over time. Latency measures how long it takes to move a small piece of information and get a response.
Written communication generally has high throughput but high latency. This is obviously true for things like physical letters in envelopes, but tends to be true electronically as well, because people tend to wander off and do something else instead of waiting for a response. So even some short conversations can extend over days, months, or even years.
One common response to this problem is to try to use higher throughput to compensate for latency. Instead of saying just one thing, people make long, structured arguments, explicitly defining terms and anticipating counterarguments or questions instead of waiting for them. In other words, they try to take the conversation as far as they can with a simulation of the other person inside their heads.
In cases where the questions or objections are easy or simple ones, this is effective - it is a convenient shortcut with a long and glorious tradition, dating back even to the days where such arguments were communicated by speechmaking and not writing at all, for example in politics and other adversarial environments where one could not trust one's interlocutor to ask fair questions and work with you to get to answers constructively. But for the hard questions, people just end up talking past each other, and have debates instead of conversations.
Good Conversation Takes Practice
This is especially problematic because it increases the opportunity cost of difficult conversations. Easy conversations get cheaper with writing (where the potential throughput is basically unlimited), so we have more of them - but the hard conversations are almost no cheaper at all by comparison. So we have very few. After all, the difficulties you have with a novel concept may be very different from the difficulties I have with it, requiring conversations that go in totally different directions, or at different speeds, or examining different parts of our vocabulary. Because of this, even if you do manage to make the points I need to hear, it doesn't necessarily scale up well - republishing the original won't reliably communicate the same thing again.
But wait - it gets worse. Good conversation about difficult things takes practice. Most people are never properly trained, because proper training is expensive and the benefits are unobvious, so they don't know what to do when the opportunity arises to learn something difficult - and instead just try to have a debate, linking to articles, citing research, making long structured arguments and explicit definitions, and trying to anticipate counterarguments before they come up. If they've started out on the wrong track, it's exhausting for even a skilled conversational partner to apply the brakes, especially because someone trained in the art of good philosophical conversation is specifically acculturated not to try to exert a disproportionate influence over the conversation.
My hope is that simply making more people aware of this failure mode will help them avoid it, but I'm not very confident this will help.