> ## Content Index
> Fetch the complete content index at: https://mereorthodoxy.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# To Efficiency and Beyond: A Warning for AI, Research, Writing, and the Rest of the World
- URL: https://mereorthodoxy.com/to-efficiency-and-beyond-a-warning-for-ai-research-writing-and-the-rest-of-the-world/
- Published: 2026-10-23T11:00:00.000Z
- Updated: 2026-10-23T10:59:59.000Z
- Description: When we use artificial intelligence in ways that short circuit our own formation, we deprive ourselves of an opportunity to grow into our distinct callings.
- Author: Mere Orthodoxy
- Tags: Jordan Steffaniak, Technology, Formation

Artificial Intelligence (AI) continues to grow, improve, and expand our horizons. It also continues to degenerate, hallucinate, and spawn a luddite every time it deceives once again. Some have fully embraced it and revolutionized their careers or family lives, finding AI can do literally everything from video generation to meal planning. Others become more deeply entrenched in their fear and distrust of AI. 

Skirmishes abound. Do we use it? For what and how often? Will it diminish our humanity? Will it enhance it? All excellent questions worthy of debate, reflection, and consideration. 

In this short essay I take up a portion of this query and sketch a brief cautionary word for the future of AI as it relates to research and writing, along with a brief suggestion on how we might responsibly use it, if we dare.

## Defining Purpose

Let us begin. Despite what the warring factions may insinuate, AI isn’t evil (though I often joke that the algorithm is run by demons) nor is it [a neutral tool we can use for unending good](https://mereorthodoxy.com/one-tool-to-rule-them-all/) (it may well be like the ring of Sauron). AI is a tool. Neither pure evil nor pure good. All tools have specific ends or goals for appropriate use, and all tools have a range of potential misuses. 

The common example is the hammer. It has the goal of driving nails through material such as wood. It has the potential misuse of hitting my thumb as I try to drive the nail into the wood or being wielded as a weapon against others. 

Or consider the scale which is designed to accurately weigh objects but could be misused by being deceitfully adjusted to create a false balance and cheat others out of what they may own. 

AI is likewise a tool that can be used for good or ill. Its primary goal is efficient intellectual production. It is a rational instrument aimed at extending human cognition and capability. But AI has numerous potential vices for the user, ending in complete intellectual dependence and degradation. While I may become dependent on tools such as a calculator for completing computing tasks I once knew by memory, my skills are not wholly eroded because the calculator has very clear limitations. Whereas AI is boundless and can erode our ability to reason in a far greater sense—indeed it can completely replace our reasoning capabilities. Therefore, AI must be handled with caution, care, and patience lest it be misused and enslave.

In order to see the danger of AI as it relates to research and writing we need to think carefully about the goal of these two activities as well. Research is for the sake of writing (usually, in this more academic context of which I currently write), though it likewise has a purpose for growth in knowledge, fulfillment, intellectual formation, and the like. Writing is for the sake of communication, persuasion, and similarly, intellectual formation. There are certainly many more goals, but these are three important ones. 

Given these goals, alongside the goals of AI, it should become clear that utilizing AI can quickly stunt our ability to achieve these ends. We will lack communication, persuasion, and intellectual formation without traditional research and writing. Why? 

Consider how we best communicate and persuade for a moment. I know precious few who would say we better persuade when we rely on technological tools for our material content from which we persuade. Instead, we best persuade when we deeply know our subject material intimately. This only comes through deep personal reflection, the sort of reflection that requires reading entire essays and books with our own two eyes rather than relying on summaries generated by someone else. 

This danger isn’t new, of course. I remember in graduate school the temptation to pretend I had read everything already which led to the temptation to rely on a quick summary (remember Spark notes?) to feign mastery of the subject. All the easier when we have a tool that can generate robust summaries of our queries in mere moments! While intellectual formation may not be at the top of the list for those who think about the ends of writing, it should be near the top. For centuries the practice of writing has been central to growing in knowledge, wisdom, and even virtue. Writing is also a form of knowing in itself. By the very act of putting words on a page our brains know with greater clarity and depth. Therefore, outsourcing this work to someone else, whether AI or an assistant, is problematic for achieving one of the chief ends of writing.

## The Dangers of Misuse

Given these ends, it should be clear that AI can severely deform the goals and ends of research and writing. It has unique dangers for hopeful academic writers, early career writers, and late-stage elite writers. Not all are the same and each ought to be outlined.

For the hopeful academic writer, AI has the massive danger of eliminating the formative process of becoming a good researcher and writer. For those not properly formed and trained yet, AI will likely eliminate the necessary formative process of becoming wise and virtuous because it is the very process of reading, writing, re-reading, and re-writing that forms the wise researcher and writer.

Permit me a brief story to illustrate. During my PhD my advisor had me write a short essay on an idea Peter van Inwagen had written on in a perennial essay of his. I wrote it up. Easy! But when I met with my supervisor to review it, he quickly noted several key areas that I was misunderstanding van Inwagen that led to critical errors in my analysis. His advice? Read it again. 

We repeated this exercise. Better. But the same end result came: read it again. More closely this time. After four to five more reads I began to truly grasp his logic and argument. This experience taught me a number of things. 

First, never make sweeping claims based on only one read of a document. Second, read slower. Third, read important works or section multiple times. Fourth, the only path to true mastery is laborious. It requires careful, slow, and painstakingly sharp reading and reflection. There is no shortcut. This is why I’d rather have a human that’s carefully read only a few documents than someone who has read hundreds, even thousands, but relied on AI to provide it. The AI bound person is intellectually captured.

And so I don’t care who wrote the summary for me. If someone else did, AI or human, and I’m responsible for explaining it or sharing it or knowing it, I’m compromised. I didn’t “do the reading.” And if I’m away from my trusty device to provide more information, I simply can’t defend myself. Nor can I truly persuade anyone. Because persuasion is as much about facts as it is about emotion and depth. Depth that is only created through time with sources.

It’s analogous in a way to two men seeking to enter a strongman competition. The modern strongman competition has a wide range of activities to test one’s strength such as the log press, vehicle pull, farmer’s walk, atlas stones, and so on. For the sake of illustration let’s assume one of these men, in his training, performs these activities alongside standard lifts such as the traditional squat, deadlift, bench press, etc. while the other man focuses primarily on machines such as the leg press, Smith machine, or calf raise. The man lifting via machine likely will lift significantly more weight throughout his training due to the aid of machines. But everyone knows the machine lifter will fail miserably in competition. Why? Because the machines artificially inflate one’s strength and remove the necessary training of muscle stabilization and the like. The man who trains with free weights will be strengthened to display his strength in any situation whereas the machine captive man will not. He can lift herculean amounts only if the environment is carefully controlled by a machine. And the true strongman will be prepared to use a machine leg press any time he pleases simply because he already has the strength. It’s not to say the leg press is bad. It’s a fine movement that may build some measure of strength. But it will never prepare one to be a true strongman.

Therefore, I argue that while AI does have a potential end that could assist research and writing, it must be wielded properly, at the right time, in the right way, in the right context, as Aristotle would encourage us. And even after such training, I argue that only a few can properly wield it in these appropriate contexts after much training in wisdom and virtue. Most will never be ready to wield it without damaging themselves and others. Most will never be provided the riches AI promises, even if used appropriately.

To use a less contested example, consider the data analyst (who in a general sense has been using some form of AI for decades, if defined properly). Most are content to allow AI to sift large data sets, assist with data analysis, and recommend forecasts and models, etc. The problem is that if the data analyst isn’t properly formed first (1) they are useless (why not cut their salary and just let the machine do the whole thing for a fraction of the price!) (2) AI will destroy the data. AI is like a very proficient intern. It needs constant direction, guidance, and checking. Things you can only do if you actually know the material yourself which can only come from laborious reading, thinking, and writing. So, while I’m certain there is a place for AI in the writing and research *process* (not the end product, to be clear. I believe writing is a characteristically human activity because it reflects rationality that is gifted to organic creatures and not sophisticated artifacts) there is precious little space for the hopeful writer.

But AI also has danger for those that have gone through the fires of formation. While AI has the potential to revolutionize our access to sources by transcribing, translating, and identifying relevant sources, it has its own unique risks even for masters. Some of these benefits we found with digitization and the internet, but with AI they are on steroids. We now have access to so many sources that it would make figures of the past likely faint in disbelief. 

But even these good ends can lead to problems of mass consumption wherein thinkers no longer have the proper depth and expertise that can only be won through hard work, patience, and painstakingly reading dense material for yourself. Some may feign expertise with the aid of AI, but when reality comes, the rest will know. The intellectual giant will become the intellectual pushover. It doesn’t matter what you say, nothing replaces time with the sources themselves. No manner of efficiency can replace it.

Those that begin to overly rely on AI remind me of the 2002 film *The Emperor’s Club*. Without giving away the entire plot, the movie culminates in a school trivia competition with the once poor student Sedgewick Bell being granted a spot in the competition through much hard work. But during the competition it becomes clear that’s he’s cheating to win. He’s being fed answers to the questions via an earpiece. The contestants don’t realize. Neither does most of the audience. But the teacher who is leading the queries does. And he feeds a question to Sedgewick he knows he can’t answer. He loses. And he never learns his lesson as he repeats this debacle decades later. Analogously, many who claim expertise today are being fed information from Claude or Gemini. They’re actually enslaved to the machine for what they think is knowledge and without recourse to the machine they are nothing but actors, hoping others buy the charade.

AI can also do the work of a dozen interns or research assistants in a fraction of the time. What a wonder for top tier researchers to continue their own work but more expansively and efficiently. The danger for them is not so much in losing their ability to research and write but in losing their ability to discern. Our discernment skills dull with AI in part because of its vast power and breadth. Who has the time or energy to fact check it for hallucinations at every turn? But over time it builds trust that goes beyond the trust earned by a premier research assistant because it becomes an omnicompetent assistant in areas well beyond the scope of the master’s own expertise. 

Where a research assistant would be able to flag a lack of competence, AI knows no bounds. Where a master would know areas of suspicion because of decades of experience, he has no such skill in areas beyond his ken. Even with carefully crafted prompts to AI to flag areas of ambiguity or concern for further research, the master is prone to forget in ways that are not so easy to forget with humans. AI has no body language to remind us of failure or incompetence.

## Conclusion

I conclude with a test-case use of AI to assist, even in this essay, in a specific way for a specific end that I hope is not vicious. I wrote this essay myself but for the sake of this topic, I asked Claude about the ends of AI and its potential misuse at the final stage. Its initial answer wasn’t terrible:

> *End*: augmenting human reasoning, drafting, and problem-solving

> *Misuse*: outsourcing judgment entirely, generating deceptive content (deepfakes, fraud), replacing rather than assisting human relationships or thought

My thesis tracks this very potential misuse: outsourcing rationality. For the hopeful writer, outsourcing rationality leads to an inability to develop the proper skills to even be capable of thinking rationally. For the mature writer, the danger of outsourcing rationality leads to a dulling of skills and potential reliance on false information that wouldn’t otherwise be tempting. The other misuses Claude lists here are real dangers. Indeed, the writing process ought to be communal. And when we lose our friends and mentors insights, we become slaves to the insights of the algorithm rather than the genius and creativity of man. 

But had I asked AI to deliver these insights to me at the beginning of my intellectual exploration of the topic I would have been intellectually captured myself. My insights would only follow the paths of the algorithm rather than the wilderness of the mind. And it is the vastness of the human mind that is the beauty of research and writing that no machine, intelligent or not, can replicate.

In conclusion conclusion (yes, double conclusion, just like your local Baptist pastor), I have two parting words: 

(1) The pioneering years of AI are at our door. The path of wisdom and virtue is narrow. The tool is capable of great good and great evil. Only a select few will manage AI well in our time. So, by my lights, it’s usually best to simply be less efficient (the horror) and not lose your soul. So what if the one ring to rule them all could conquer worlds and bring great harmony and peace. We know the depravity and instability of man. Though for the right individuals, opportunity may await. 

(2) AI will never write an excellent essay. I’m sorry if this hurts your feelings. But it’s true because excellent essays *really* think and feel. Two things that are intrinsically human and organic. So, if you want to actually be a researcher, writer, or thought leader™, you’ll need to think for yourself. You can already see the proliferation of AI generated resources out there and they all have the same tells. People know. Because we’re created like that.

Those screeching about AI being amazing because it has limitless opportunities, levels the playing field, destroys the so-called intellectual gatekeepers, and transcends human intellectual capabilities are usually those wedded to it in an unhealthy way. They are believing false promises and projecting their own anxieties that others would find out their true lack of intellectual prowess. So, dear reader, remember also that the end of humanity isn’t efficiency. It’s truth, goodness, and beauty. Efficiency can lead us away from this end. The god of efficiency is exhausting. It’s a cruel task master. It’s deeply modern, destructive, and psychologically paralyzing. Follow the path of those of our glorious past and hopefully future. Pursue truth, goodness, and beauty in all its forms, but especially by reading, writing, and conversing with real people along the way. You’ll be formed in true wisdom and virtue that way.

## Appendix

For those unwilling to embrace my warning and who dare to explore the capabilities of AI in a virtuous way, I do want to offer a few concrete considerations for AI in research and writing that may provide a way forward. After all, I’m what they might call a Zotero evangelist. 

I believe the tool is incredibly useful for saving time and allowing the write to focus on what matters most—researching and writing. Not fixing the footnote for the tenth time as you re-arrange your prose. Though there is something to be gained from becoming a true master of formatting, I believe the rewards of focusing more readily on the actual research and writing itself is of superior value. But Zotero isn’t exactly AI. It’s a generic technological tool for research and writing. True enough. Given this, I might suggest using resources like Claude for the following.

First, AI can assist in research compilation. When I first started in my PhD, my supervisor had me scan every major journal issue for my thesis topic and create a list of all the essays I needed to read. That took forever. The scanning of abstracts and titles did not aid me much (besides the curiosity of finding off-topic essays I wanted to then read). It could easily have been outsourced to a tool that could scan all the essays on my behalf for certain keywords, just like I did, allowing me to have a more comprehensive list, faster. Furthermore, I no doubt missed certain essays because of my own exhaustion or hasty skim. I’ve tested this for myself several times. I’ve asked Claude when I initially begin researching a brand new topic: “Find me all the most important resources on this topic.” It usually is capable of finding several of the keys works though it intermixes irrelevant or useless resources at times. But this isn’t outsourcing the actual rational process, in my understanding. It’s hastening my ability to explore and discern for myself.

Second, AI could function as a sort of Devil’s Advocate. We could let it criticize our core argument and suggest areas of weakness. This is a dangerous route, of course. Many are quickly tempted to let Claude do *all of the work* at this stage. But for the wise man, this can be of great value. Exposing weaknesses earlier in the process and allowing for greater depth of exploration. Though it cannot and should not replace interaction with real people in real time. People provide truly original criticisms that AI never could imagine. In this way, AI is a handmaiden, kept in its proper place as research assistant.

The problem here is when we are constantly pushing the boundaries of acceptability. It will lead us to thinking we can wield the one ring to rule them all. To quote Arrested Development: 

> Lindsay: “Well, did it work for those people?”

> Tobias: “No, it never does. I mean, these people somehow delude themselves into thinking it might, but... but it might work for us.”