I offer you a close reading of a single line of text that startled me as I was perusing a seventeenth-century educational treatise.
I am sitting at a writing retreat, drafting a research paper. Those who know me would be unsurprised to learn that the PDF open on my screen contains a work by John Amos Comenius, a long-term dialogue partner of mine. But the line that has snagged my attention is not in his treatise. It appears in a superimposed prompt from the software, apparently an attempt by Adobe to help me work the right way. The first sentence makes me smile:
“This appears to be a long document.”
No kidding. Thanks for the insight. I wouldn’t have noticed. (I do appreciate the cautious avoidance of dogma about what exactly constitutes “long.” Comenius was not assuming twenty-first century attention spans, but the work in question is hardly War and Peace.)
It continues:
“Save time by reading a summary provided by an AI assistant.”
This is somehow both more and less amusing. More amusing because the idea that an intricate argument from another century would be understood better by substituting an AI summary is foolish beyond words. Less amusing because the idea that I would want to substitute a summary from an inappropriately personified “AI Assistant” is sad, the presentation of that option in the imperative mood is disturbing, and the surface plausibility of such a suggestion within our present social imaginary is depressing.
Of course, I remind myself, the algorithm is not based on knowledge of what I am reading or why it matters. Its designers probably did not intend to trash the writings of one of history’s great Christian thinkers, or to insult my own attention span, level of interest, or willingness to care. The prompt is just an attempt to provide tools that might make it easier to cope with a paralyzing flood of PDF input, most likely of an ephemeral and bureaucratic nature. There are no doubt legions of documents forming part of many people’s burdens for which a concise summary would be a valid and welcome salve: meeting minutes, reports, survey feedback, and the like. I also grant that even with great literary texts, a good summary has its uses, perhaps to get the lay of the land before deciding where to focus (Comenius wrote a LOT). The summary need not be a way out. If we mind the dangers of creating initial bias, it could be a precursor to more charitable, just, patient, careful reading by a finite creature with limited time and other real responsibilities.
As I thus talk myself into a charitable disposition toward Adobe’s interface designers, it occurs to me to wonder whether the folk at Adobe have thought about which kinds of documents might be well served by an AI summary. It turns out they have. There is an Adobe support page explaining “the most helpful types of documents to use with an AI summarizer.” The very first suggestion is educational assignments:
School comes with a lot of textbooks to read and information to absorb. If you’re running low on time, upload a section of your textbook or study guide as a PDF, Word document, or PowerPoint to an AI summarizer, and almost instantly, you can receive high-level learnings provided by the Generative Summary tool in Acrobat. This makes it perfect for last-minute study sessions or adding to your notes.
By now, I am thinking that Adobe and I are really not on the same team. I am also thinking that before judging the laziness of students who do not do the reading, I should note that they may be following the instructions that come with their tools.
As I labor to remain charitable toward the Adobe team, I will allow that there are moments, such as reviewing earlier work just before a cumulative exam, when a good summary might be a useful part of a viable learning strategy. Adobe’s marketing patter has none of that nuance. They simply recommend using the summarizer “before class.” (The bizarre notion that a third-party summary amounts to “high-level learnings” is a further can of worms that I set aside for now.)
Back to the prompt on my screen. What seems most worth pondering is that it is indeed a prompt. It is a pedagogical device, an invitation to learn to do things differently. My software is setting out to teach me something, and like all acts of teaching, this one is far from disinterested. It invites me to share a specific set of values and assumptions. Prominent among the values and assumptions communicated are the following:
- That I am most likely reading text in which the detail does not matter to me or anyone else. The prompt does not create a conditional (“If this is a….”); it just tells me how I should deal with whatever I am reading if it happens to be long.
- That this is the normal reading state, that everyone else is just summarizing disposable text, and that I am odd if I insist on doing otherwise. The prompt is a default; it projects the assumption that everyone wants to “save time.”
- That my most fundamental value is efficiency (not, for instance, care, justice, responsibility, humility, charity, or patience), that “long” therefore means “bad.” Conversely, the most fundamental danger is having to read unnecessarily (not, for instance, the danger of shallow understanding, missing important details, failing to grow as a reader, lack of insight, evading patient reflection, erosion of attention span, or missing the Spirit’s whisper). The prompt does not hedge (“If you are short on time…”). It tells me which problem I have.
- That the point of reading is task completion, not careful internalization. The prompt implies that success means moving the text out of my workspace faster and that the summary replaces the text. (If careful reading still ensues, time is not “saved.”)
- That if I fail to share these values, I am falling behind, and that moving faster is the way to win at life. The prompt suggests that I need help shifting to a better way of doing things and that I should catch up with the cool kids and power users. (Of course, this is also intentional marketing pressure; clicking the banner leads eventually to a “subscribe” button.)
- That the point of education is to have a summary of the main points (whether or not you understand the process behind it) and look as if you engaged with the material, and the point of a text is its key assertions, not the texture, grounds, or validity of its argument.* (This applies more to the support page on the Adobe website and is as much an indictment of schools and teachers as of students or software.)
I am sympathetic to the challenge of writing concise software prompts. But we can imagine a world in which such prompts would be different. “If you need a summary, try…” might be a small step in a more helpful direction. Rewriting the Adobe help page with actual education in mind might be another. Yet what stares at me from the screen is an imperative to prioritize speed over understanding.
Of course, quixotic as the effort may be, I am trying to turn the tables here. I have taken a succinct, efficient fragment of consumer AI culture and turned it into a much longer text, attempting a slow reading in a quest for insight. Comenius, my long-term dialogue partner (a concept little imagined by the Adobe prompt), lamented our drive to turn formation into information and to prefer snippets over internalized wisdom. The printing press was his disruptive technology. He recommended reading fewer things well. If that is wise advice, then our struggle is not primarily against student vice, but against the commercial and educational construction of a culture of engagement with texts and ideas that is most deeply shaped by our technological choices and speed-soaked values.
I’m afraid this has become a long document, an unnecessary elaboration, a slowing of my time and yours. I’m sorry. To save you time, I prepared a summary:
The tools that we and our students use daily are not just tools; they are pedagogical devices. They bid to shape how we learn. Your software is telling you how to read. If efficiency, good as it is in its place, is in fact not the most fundamental desirable value, the software may not be forming you or your students well. You need to actively and intentionally discuss with your students not just what to read, but how to read it, and which prompts to ignore.
* A Final Note for Determined Readers.
Naturally, curiosity overcame me, and I had to try running the above reflections through the Adobe summarizer. As expected, it summarized the bare bones adequately, but it also translated the ideas into current conventional language. It appealed only to the need for “critical thinking,” a common if often vacuous term never actually used in this essay. It did not pick up on any of the repeated references to ideas such as charity or justice. It missed all hints at a Christian subtext. Perhaps you did better.
