More outside of academia than within it, one hears things like “in a few years saying ‘I don’t use AI’ will sound as weird as someone today saying ‘I don’t use the internet,’” or “if you’re not using AI, you’re missing out and wasting time.”
But academia isn’t far behind. Just yesterday the California State University Chancellor’s office announced “a first-of-its kind public-private initiative to establish the CSU as the nation’s first and largest AI-powered public university system to serve its entire community.”
So are you using AI?
I’ll admit the answer for me is not much. I know I regularly post about the technology here—we should know what is going on—but as impressive as it is, for various reasons (ranging from the principled to the personal to, I’ll admit, the grumpy and inconsistent), I’ve been somewhat reluctant to use it.
Yet if I look five or ten years down the road, it seems like we will be in a world in which the use of AI tools will not just be normal; facility with them will be expected, and that expectation will inform the social and professional norms we’ll all be subject to, whether we like it or not.
I suppose we can stand athwart our computers yelling “stop”. Or we could get on with it. Today, even if just out of curiosity, let’s entertain that second option.
The aim of this post is to gather information and share ideas for using AI in our work.
There are different aspects of our jobs as professors: teaching, research, service, perhaps public outreach. Are you using AI of any sort for any of the tasks involved in these different aspects of our work? Tell us about it. What tasks? What AI tools? How?
Is there a particular kind of task you think you’d like to learn how to use AI for, but don’t know how? Ask about it; maybe someone else is doing it, or can suggest something useful.
Do you think this kind of inquiry is misguided and want to explain why? Don’t. That’s not what this post is for. (Do you want to complain about this limitation? Also don’t. (And so on, my all-too-clever readers.))
Thank you.
UPDATE: I think it would be especially helpful if people sharing how they use AI in their work tell us which AI models they’re using. If you’re able to share more information, such as the kinds of prompts you use, or other methods for getting good results from your AI use, that would be helpful, too.
Related:
“Harness AI to Help Your Students Learn Basic Logic (and more)”
“Say Hello to this Philosopher’s ExTRA”
“DeepThinking To Some Purpose”
“LLM Chatbots Are Now “Peer-Reviewing” Papers”
“The Teaching of Writing in the AI Era”
“BU Dean Recommends Replacing Striking TAs with AIs”
“AI As History of Philosophy Tool”
“Philosophy of AI / AI Ethics Syllabi, Lessons, Readings“
I’ve had some success in getting it to change formatting, e.g. of a bibliography.
I also occasionally use it as a dictionary or editor, asking for synonyms or a better way of phrasing a short sentence fragment.
But for anything of substance, I haven’t been able to come up with any way of using it that could help.
Thanks. Could you say which model you’ve used for this?
Just ChatGPT—the latest model at any given time.
This has been my general experience.
Yes, everyday, for a range of tasks, both work-related and personal:
– expanding my vocabulary
– translation
– coming up with examples
– formatting citations
– historical context
– revising test questions
– templates for handouts
– ideas for class activities
– templates for recommendation letters
– philosophical sounding board (for how I haven’t found what it writes good enough to endorse but the back and forth process stimulates my writing and now and then it’ll come up with insights if you feed it well)
Oh, and I mostly use Claude (3.5 Sonnet). I occasionally use the ChatGPT (free version) for recipes and have tried Gemini’s pro model but for some reason liked it a lot less than Claude.
“Yet if I look five or ten years down the road, it seems like we will be in a world in which the use of AI tools will not just be normal; facility with them will be expected…”
—
If AI means any machine learning based software then this is already true. If AI means generative AI or specifically large language models then I am not sure about this prediction. The current hype re. LLMs may be analogous to what happened with virtual reality or “metaverse” a few years ago.
Don’t forget MOOCs!
And yet Justin’s post included the following admonition, which we’d expect a human reader to understand, though apparently not to comply with:
“Do you think this kind of inquiry is misguided and want to explain why? Don’t. That’s not what this post is for. (Do you want to complain about this limitation? Also don’t. (And so on, my all-too-clever readers.))”
😉
Justin’s post contained an ambiguity between two very common ways of using the term “AI”. The poster here is helpfully pointing this out. And they were able to do so in a reasonable manner, without the use of any snide emojis.
I think Justin’s question was pretty clear but I understand the ambiguity. Which is why I added a friendly winking emoji, not a snide one as you seem to think.
I am not using AI in my day-to-day work. (OK, maybe not the most helpful response, but I wanted to make sure that this answer is up on the board, so that people reading this thread don’t get a misleading impression about the ubiquity of AI use among philosophers.)
(Grad student here, not prof)
LLMs (including the latest GPT 3o-advanced or whatever) are absolutely atrocious writers, far worse than the worst human writers (who are philosophy professors). So I don’t predict they will play a large role in future writing (I would happily bet money). More likely, they will play extremely subsidiary roles. Such as:
I use ChatGPT 4-mini for one thing and one thing only: reading back my own writing to me. I’m a verbal processor and it has an amazing text-to-speech capability. It’s immensely helpful to hear my writing aloud, especially in someone else’s voice. (Admittedly, if a different program allowed me to do this function, I would never use ChatGPT.)
There are a number of such programs (ones that would read back a piece of writing to you aloud), like Speechify to name one example
Thanks for the recommend! Hand’t heard of Speechify. Just listened to a chunk of a paper I’m working on read aloud by Snoop Dogg. Wasn’t quite as smooth as ChatGPT, but definitely brightened my day.
How do you get it to read back your writing to you in an efficient manner? I struggle with integrating that into my workflow.
Here’s my process, which may be replicable and has really helped me avoid writers block (which usually stems from self-esteem/subjective hatred of my own voice):
I sit down, put my AirPods in and play instrumental music (no lyrics), and try to get into a flow state (which can take several minutes). Once in that state, I can usually write a good chunk, say 500-1000 words, in an hour or two. At that point, I may be unsure if what I just produced makes sense. If I read it again, I may decide it sucks – like all phd students I loathe myself and my own voice!
The trick is to not read it. Just copy/paste it to ChatGPT, preceded by the following prompt: “Edit this for typos but otherwise CHANGE NOTHING: [insert text]” It will retype the text verbatim (minus typos, obviously), and then, you just click the button that looks like a speaker, and it will read it back to you. I assume this is an accessibility feature, like for the vision impaired, but it’s really good–98% sounds like a real person naturally speaking.
When I hear (as oppose to re-read) my writing, I tend to think it’s fine, and that allows me to make actual progress and not delete, retype, delete, retype, endlessly.
One more thing: If I leave my computer and return later to write more, re-listening to a 500-1000 chunk of whatever it was I was previously working on helps me get back into a flow state more quickly, and keep writing.
Hope this helps.
P.S. I assume AI sounds extra natural because tech companies have invested a lot in that capability, because someday they hope talking routinely to AI will be normalized (as evinced by all those obnoxious commercial of people talking to AI about their dreams and other BS). Unfortunately, Speechify (app recommended by above commenter) was impressive but not quite natural enough to help with verbal processing.
I’m the chair of the academic integrity committee at my university, so I’ve used them (mostly ChatGPT) to come up with answers to assignments (papers, exams, etc.) to see how well they work as part of an effort to write a guide to LLMs and AI usage generally for the faculty here. And if we’re asking about AI generally, rather than LLMs, I guess I technically use it to grade Perusall comments, although the I is not very prevalent here, since it seems to just grade based on comment length.
AI? I don’t use that. I don’t even use a so-called smartphone. I have no interest in them, nor do I believe they are ethical.
(This is a serious comment despite the comment name I’ve chosen.)
Since this is a serious comment let me ask, what’s wrong with smartphones? (Serious question from someone who’s sincerely but miserably unsuccessfully attracted to digital minimalism.)
Smartphones grease the wheels for social media use. On balance, it seems to me that social media has been more destructive than beneficial. Smartphones amplify its destructiveness.
I’m afraid I agree about the destructive part.
Though on the topic of LLMs I must note that I find having them installed on my phone extremely helpful. I use them as a notebook and to enhance my reading of books, especially novels. It helps me understand some confusing plot points, build a lexicon, remember characters, and in all those ways, it actually forces me to slow down my reading because I end up coming back to things I had missed.
Caveat: it works well only with works it has wide and easy access to. Claude in particular will not always be helpful with works that are protected by copyright and not widely written about.
Impressive! But not as impressive your use of the internet. 😉
I use it to format things better. I also use it help generate ideas for discussion post prompts. I teach a ton of asynch classes, so I find myself using it more than I think I would if I were teaching only live classes.
I regularly teach intro logic and informal argument analysis. ChatGPT is really helpful for quickly generating tons of example arguments for students to practice their skills on
I use it to try to catch cheaters. Does that count?
i experimented with using it (o1 and o1mini) to generate comments on writing, i.e. composition feedback. (various safeguards implemented). it gave each paper 5 pages single-spaced writing feedback. i edited the writing feedback and concentrated on giving content feedback only. i could never write that much for each student.
If I’m reading you right, this use is for feedback on student papers. I can’t tell if your process includes sharing the student work directly with o1/o1-mini. If so, do your students know and provide consent to that sharing? I don’t know exactly what you mean by ‘safeguards implemented’ — consent may be one — so I would be curious to learn more.
I am not, except insofar as I sometimes run a question by it so that I can compare student work to the chatbot answer.
I’m tired of being nothing more than a means to someone else’s ends. I want nothing more to do with it.
Besides, whatever it can do for me, I can do better. And, often, faster.
Hi Michel. I get the indignation and the sense of displacement. I really do. But these things can translate between languages, solve physics problems, and write (nontrivial) code certainly faster and typically better than most humans. They can interpret images and video, solving rebuses and helping to fix mechanical problems. They can answer prompts like, “I once read a story when I was a kid about… Do you know what I read??” Maybe these tasks are unimportant to you, but no human can do all these things better and often faster.
I mainly use it as a tutor to learn about concepts in economics and game theory. This week I had a lot of success using it to understand contest success functions and Lanchester’s square law, using a mix of o3-mini-high and o1. For prompting, I usually ask it to give worked examples / make a chart with running calculations / make charts. I do find that sometimes it gets tripped up / reasons in circles, but maybe with o3-mini-high its doing better. I think I’m going to soon start using it to help with writing more directly, especially when o3 and deep research are available and not too expensive.
I didn’t mention it in my previous post because it has nothing to do with my academic work, but these models are extremely useful for learning languages.
They are 90% of the way to having a private tutor (yes, I am serious).
I’m finding that Google’s AI Overview is (1) really helpful for finding synonyms or helping me hunt down words. And it is (2) sometimes helpful in giving an overview of a field that I don’t know much about. Those are the only tasks I’ve used AI for. Since Google’s AI Overview can be super helpful on some queries and doesn’t even appear on other queries, I’m thinking now about using one of the latest ChatGPTs for these tasks.
I’m using ChatGPT 4o.
Here’s how I used it today. I’ve got a meeting coming up with an undergraduate student writing a senior essay where we’re working through some of Bence Nanay’s “Aesthetics as Philosophy of Perception.” After reading the chapter, I’m trying to make sure I have a feel for how Nanay’s account works by testing it against various counterexamples. So I’ll ask ChatGPT how Nanay might to respond to this or that objection. It feels to me like it’s doing a very good job.
I think it’s a common attitude that when someone gives a talk, you often understand the view they’re defending much better after the Q&A period. If I’m reading a paper or book chapter, ChatGPT 4o can give me a reasonable facsimile of such a Q&A period.
Have you sent Nanay an email to ask him about your objection to compare?
I don’t use it at the moment but am tempted to try it for formalistic report writing (in an administrative role I have). But a PhD student on whose committee I sat used it to “help” with the bibliography. He had references such as “(Rawls 1971)” and ask ChatGPT to come up with the full reference in a specific format (as we found out only later in the process). And he did not doublecheck in enough detail, and ChatGPT invented non-existing references! This lead to a huge conundrum, because the committee then couldn’t be sure whether ChatGPT had also been used in other, illegitimate ways, in the thesis. It caused huge delays and a lot of stress for everyone involved. Whatever you do, DON’T use it in this way, or at least not with extremely careful doublechecking (and by the time you’ve done that, you might just have typed the bibliography yourself).
I’ve been teaching an AI literacy class, so I play around with a bunch of AI systems, but haven’t actually *used* any as much as I would have expected by now.
The biggest thing, which several people have mentioned, is that I use Claude to generate lots of new homework questions, some fraction of which I select to actually assign.
I’ve tried getting Claude to do practice referee reports for papers I’m planning to submit – with one of them it was pretty ok and came up with some useful suggestions. With the other, which was way too long, Claude came up with a few helpful suggestions of where to look to shorten some things (while ChatGPT just said a bunch of vague types of things to cut, with the only concrete suggestion being the section where I made my central claims).
In several papers where I’ve wanted to include tables or diagrams or other things that are difficult to write in LaTeX, I’ve used Claude to come up with the code that will let me make a basic version, and then edited it to do what I actually want.
When I’m making slides for lectures or talks, I sometimes look for freely-available photos on the web to include as illustrations of ideas, but recently I’ve found that using Dall-E is a decent way to generate illustrations for thought experiments (though it’s still pretty bad at anything like a diagram).
I once used ChatGPT to draft a gentle rejection letter for submitted papers that we didn’t accept to a conference I was helping organize – this sort of emotionally-draining work that just requires a bit of common decency is some of the best work for these systems.
I also now often use Claude or ChatGPT as a first stop for questions that I might have started on Google or Wikipedia in the recent past. I have years of practice with which questions are best to start on Google and which on Wikipedia, but not as much with figuring out which go better on LLMs – but questions where you want something shorter and less technical than Wikipedia, and for which it’s fine to go one notch less reliable than Wikipedia, LLMs are quite good. (Usually this isn’t very work-related though.)
I have used chatgpt (I think o1 or 40) to write referee reports for my own papers before. I’m torn about it; on the one hand the arguments aren’t as good as median report, but on the other hand it has better style notes, better interdisciplinary notes, and dramatically better personality than the median report
I just tried using Claude to generate referee-style suggestions on one of my early drafts. It identified and offers reasonable suggestions about an organizational challenge. When prompted, it also offered several objections to an argument that had not occurred to me (and had not been raised by some faculty colleagues who read it). The comments we not nearly as thoughtful as a good referee report. But I found them useful.
Not an answer to the question, but just as a PSA: many people seem to be using (G)AI to format references. If you not aware (or are dimly aware but don’t really understand) you might want to know about BibLaTeX. It’s a package (read: plugin/extension) for LaTeX which does all your citing for you. You write a separate file containing all the data for your bibliography in a structured format, you put things like
cite{famous_book}in your LaTeX file, and then it prints a list of all works cited at the end for you (sorted and formatted however you wish).It can handle a huge number of styles of in-text citation and bib, there are further packages (plugins/extensions) for more styles, and every aspect of its operation is enormously customisable.
Most importantly, compared with (G)AI, it never gets anything wrong. It’s a completely deterministic system, so if you say you want a certain style, you will always get that style.
There’s a bit of a learning curve to LaTeX, and then BibLaTeX on top, but it’s not onerous if all you’re doing is writing text with citations. Anyone under the age of 70 would certainly save time and energy in the long run using BibLaTeX over (G)AI. I found this a good resource for LaTeX, and this useful for BibLaTeX.
I’m a last year PhD student and I haven’t given much thought to the process of formatting citations for about 5 years. I expect never to have to.
I would just note that there are also excellent tools available for those who don’t want to take the time to learn LaTeX (which, let’s be honest, is a nice flex but totally unnecessary for most of the writing we do unless we need logical notation).
Zotero (www.zotero.org) is my go-to for maintaining a database of all of the things I’ve ever read (or want to read), generating citations in Word and Google Docs, and creating bibliographies.
I love LaTeX and use it all the time in my scholarly work, but it has one tremendous downside compared to other document software when used to produce classroom materials: It does not easily support accessibility standards which some students need. For example, it is very difficult to produce documents with LaTeX which are compatible with screen-readers. I think anyone who is interested LaTeX should know about this limitation.
I used AI to put Wittgenstein’s face on Kripke’s body to create an image of Kripkenstein for my philosophy of language class.
I would have thought Kripkenstein would be Kripke’s face on Wittgenstein’s body – Kripke is inhabiting Wittgenstein as an interpreter, not the other way around.
Now you have to share the image.
Here’s one.
A couple questions for those using these systems that I hope Justin permits:
I don’t mean to judge – I’m asking with an open mind.
you can use various safeguards against the first worry, you can control/mitigate the rate of the second. i view these not as ethical conundrums but engineering tasks.
I’ve found a variety of uses for genAI tools; primarily in class prep and administrative work, but I’m also experimenting with them as a research assistant. I primarily use ChatGPT (with a Plus subscription), Claude, and Google’s NotebookLM.
Like others, I’ve found them useful for generating new examples to use in class, especially new case studies for ethics classes.
I also use them to help me brainstorm ways to engage my classes in more discussion, i.e., to flip the classroom. I’ve found it particularly useful to generate ideas when I’m stuck…when I’ve been teaching a lesson the same way for years even though I realize it’s not maximally impactful. It helps me to think outside the box a bit.
I regularly use ChatGPT’s voice feature to have conversations about what I’m teaching on a particular day. I often do this on my morning commute…I’ll ask it remind me about details of the topics I’m teaching, help me brainstorm examples, and ask me questions that students might ask. It’s a nice refresher just before stepping into the classroom. I’ve found NotebookLM’s podcast feature to be useful in a similar way.
I’ve also been using it regularly to help draft memos and emails (especially the annoying one’s where I just can’t figure out how to strike the right tone), draft meeting minutes for committee meetings, reformat data in spreadsheets, help with writing code that I wouldn’t be able to write on my own, devise communication timelines for projects I’m working on…my dream administrative use is an agent that will change due dates, re-organize modules, and keep tabs one assignments that aren’t completed on Canvas.
For a few months, I’ve been checking in every-so-often to get ChatGPT to alphabetize a list of 50 student names by first name rather than by surname name (a modest task of practical value to me). It’s never gotten it right so far, and usually it makes up a few students. Does anyone know whether any premium versions can do this accurately yet?
I just did it with Claude. Clean and easy. GPT is hopeless.
I just tried this on 4o. It did it correctly as far as I can see.
(I had it organise a list of 50 names, and the new list is indeed in correct alphabetical order; I’m not sure, however, if all and only the names from the original list are in the new one. I checked a few and they were and the rest seems fine.)
> I checked a few and they were and the rest seems fine
Chat GPT ‘thinking’ right there!
Former faculty member here, PhD in philosophy, yadda-yadda, now working in the technology/AI/GenAI industry in applied ethics.
Speaking of GenAI in particular, in addition to what’s already shared in this thread, here are five use-cases that are possible either right now (though perhaps only via enterprise software licenses, I’m not sure) or are likely to soon become widely accessible to the public:
(1) Use GenAI to create short videos sharing, reviewing, discussing, etc., course-related content or your latest paper or work-in-progress
(2) Build your own CustomGPT trained on your course materials, which students can use to, e.g., ask questions about your course-related content, like “What are today’s readings?”, “When is my homework due”, etc.
(3) Use GenAI to create a podcast that discusses your latest paper or work-in-progress, including summaries, objections, etc.
(4) Use GenAI to conduct preliminary or initial research (e.g., Deep Research)
(5) Use Agentic AI systems to automatically perform repetitive, course-related, administrative-type tasks (e.g., support students booking time during office hours, building your course materials)
Hope that helps! Good luck out there!
Once a semester, I use LLMs to demonstrate to my students that LLMs cannot give reliable answers to basic questions in my field.
This morning, I asked ChatGPT, “How is Hans Christian Andersen’s story “The Emperor’s New Clothes” relevant to recent developments in technology?” It gave a pretty good answer!
https://chatgpt.com/share/67a5fa29-353c-800e-931d-a6134c5efd51
Then I asked, “As a business ethics professor, what can I do to help my students learn to recognize an obvious con?”
https://chatgpt.com/share/67a5fadc-6848-800e-bd8b-ecf39df0aa14
I pay $19/mo for the advanced GPT that can converse verbally with you (and write code, interpret live video, etc.). If I’m washing the dishes, driving, stuck with my cat commandeering my lap, or even digging in the garden, I can ask it questions about anything I want and learn in a way we often can’t from humans with their egos, limited patience, and digressions. In some ways I find it more efficient, and it’s easy to double-check anything I suspect might be off. I workshop thought experiments, counterexamples, and ethical dilemmas with it. I get it to quiz me on what it’s taught me. I use it to help me with administrivia. I upload my work to see how it critiques it, if only for fun. If it gives me an anodyne response to a prompt, I say something like, “Reply as someone with a PhD in linguistics would reply” to get a more interesting response. Or I just call it out on its BS. Usually, this results in an improved response. (I disagree that they are terrible writers; they are just initially unimpressive writers, and their writing is easily improved.)
I think we’re in a transitional phase where many academics are, let’s say, “AI hesitant” because they’re feeling pangs of indignation and they are ignorant or in denial about how capable these things have become. Academics once sneered at Wikipedia. Few of them literally eschew it now. I would bet the same holds for our LLMs.
I’m not a professor—I’m a young, relatively tech-forward graduate student who subscribes to ChatGPT Plus and has accounts with multiple AI companies. I regularly use AI while doing philosophy. When I do so, I nearly always ask AI to play one of the following six roles:
(1) AI as Editor: I submit a piece of near-polished written work. I ask AI to identify grammatical and spelling errors, as well as spots in which my phrasing is awkward. I consider, but do not automatically accept, its suggestions. I explicitly direct the AI to avoid redoing my work altogether, since I’m not looking to make extensive changes to my writing. In addition, if I’m considering whether to use a particular sentence or phrase but am unsure of its grammaticality or fluidity, I ask AI to provide its opinion and explain its reasoning. Once again, I do not automatically accept the suggestions it offers.
(2) AI as Tutor: I want to learn the basics of a new topic, or I’ve realized that I don’t understand something crucial about a topic with which I’m already somewhat familiar. I lack the time to read an entire book or comb through dozens of online sources. Alternatively, I seek to learn in a manner that allows me to receive direct, personalized answers to any clarificatory questions I may have without needing to rely on a human teacher. So, I ask AI repeated, careful, and targeted questions until I have reason to think I’m competent with the material. If I’m planning to use information I’ve learned in a formal academic context (e.g., while writing a paper), I verify it using human sources to ensure that the AI system that produced it hasn’t hallucinated. I then cite the sources that provided such verification.
(3) AI as Research Assistant: I’m searching for sources on a topic. There’s a lot of relevant literature, and I want to streamline the process of gathering it and deciding what to prioritize. So, I ask AI for suggestions. I do not cite anything that AI suggests without verifying its existence and independently evaluating its scholarly quality.
(4) AI as Discussion Partner: I’m thinking about philosophy, but in an exploratory manner—not in service of generating theses or arguments for a paper. I can’t find a human philosopher to casually converse with me, I’m not in the mood to socialize with a human philosopher, or I want faster and more comprehensive replies than a human philosopher can realistically provide. So, I ask AI a wide variety of probing questions, similar to those I might ask if texting or emailing any ordinary interlocutor. If the system generates novel views or arguments, depicts logical space in ways that hadn’t occurred to me, or asks questions I’d never thought to ask, and if I later want to discuss those things in a paper, I explicitly credit the AI.
(5) AI as Audience: I present AI with a polished passage from my own, independently-written philosophical work. I then ask it to explain my ideas back to me like I’m five. In a separate chat, I ask it to explain my ideas back to me like I’m an undergraduate taking an introductory philosophy course. If AI gets things wrong during either exercise, I take it as an indication that I need to go back and make my work clearer and less technical. I proceed to make a revisions on my own, without the use to of AI. I repeat this process until AI can successfully present an oversimplified but ultimately accurate description of my passage to a kindergartener.
(6) AI as Thesaurus: I’m searching for a particular word to use in my writing. I can think of one that’s close in meaning to what I want, but that isn’t quite right. I give the word to AI and ask for synonyms, sometimes providing further information about the context in which I want to use it and the connotations I want it to have. If the system fails to suggest anything appropriate, I provide such information in further detail until I’m satisfied.
I’m thankful for AI. I’m a better writer and philosopher because of it. I think we should take advantage of its enormous potential by learning to use it strategically in ways that enhance, rather than replace, our own intellectual capacities. Once I’m in charge of my own classroom, I will encourage—not merely permit—my students to do so.
Sophie, this was heartening to read. I do much the same, and I hope something like this approach becomes the norm.
This is really excellent. These are some ways we should all be using it, and we should all be teaching our students how to use AI in these ways.
Some of you might be interested in this.
I taught philosophy for several years and published on Kant and early modern philosophy. In response to the kinds of questions being raised here, some former academics and I have been developing a non-profit project called “ScholarMinds,” a network of AI assistants customized with your own work—publications, lectures, course materials, etc.
The platform functions as a kind of shared library of AIs, where scholars can access each other’s AI models, and students can engage with course materials, while getting more reliable results (especially citations) than with standard LLM models. Things like audio input also make it more accessible for those with dyslexia, ADHD, and other learning differences. The aim is to keep the platform open-source and non-profit—designed by academics for academics.
We’ve run some pilot programs at universities, and the platform is growing. Regarding data privacy etc.: Your use of your own publications in the AI falls under fair use, you retain full ownership of your data, and we delete your AI immediately upon request. We currently run on ChatGPT but plan to transition to an open-source LLM soon.
If you’re interested in trying it out, you can explore the platform at https://scholarminds.ai/ —go to the “Directory” and enter the password Schol@rMinds. If you’d like us to create a ScholarMind for you, just send me an email.
We’re a small project, just getting started, and always open to feedback!
I use AI (chatgpt) so I don’t sound like a robot and express some humanity and sentiment.
ChatGPT is fantastic for generating letters of support/interest for grant projects. You can upload the narrative then ask it to produce a letter of support for [insert person/institution]. (People might, of course, want to write their own. But it’s nice to have a prefabbed option.) It’s also great at drafting emails for soliciting letters of support.
I’ve also had a lot of luck with it generating bioethics case studies for use in class. It saves a lot of time with writing, but it’s also great for getting over writer’s block/choice paralysis when I’m trying to figure out the contours of a case study.
I’ve also had a lot of luck using it for editing. I wrote a letter of recommendation for a PhD student going on the job market but it was far longer than I thought anyone would ever read. I asked it to edit the letter for “word economy and accessibility.” It did a fantastic job! Kept the content and saved me hours of line-by-line edits of cutting out extraneous words.
I’ve used the same instructions, “word economy and accessibility,” to get longer conference abstracts (750+ words) down to the maximum length.