When you limit the Study Assistant to "Your Books" does that include books in your print library?
Good question!
It does not include the Personal Books that you built yourself. Don't know about print library.
Print library books are part of the Logos catalog, so they should be included. Smart search includes them; I would think the SA includes them as well.
Waiting on τὸ στόμα τοῦ ἵππου
You are right, the horse's mouth is needed here. I did an experiment with John Piper's Don't Waste Your Life (which I have in my print library). The first query was run on "Your books" and, although it goes on to answer the question (from general knowledge, yikes), it states that no matching book was found in my library:
The second query was run on "All books":
With this single experiment, it does look like the SA doesn't recognize print library books as part of my library, but I won't bother testing further.
@Mark Barnes (Logos) Can you confirm whether the Study Assistant is taking our Print Library books into account when answering queries? If not, is there any future plan for that? Thanks.
I don't know, but I will find out.
I have a number of books I bought from Logos in my Library. A number of them are from a non-orthodox biblical perspective, e.g., Biblical Unitarianism. The AI Assistant told me that these books are not included in its training data. It even told me that the dataset it is using is a "curated" dataset, determined not by the books in my Library that I bought from Logos, but only from the books that Logos has put in the dataset. This effectively means that any answers you get, with references, are limited to the Logos Developers' perspective, and references will be limited to those "approved" sources. Everything then becomes effectively an appeal to authority. The Authority is determined by the training data. You'll get the standard evangelical orthodox perspective or nothing. If the AI were trained on every book available for purchase from Logos, for Logos, then you would get references from many perspectives, even unorthodox ones. One should be able to limit the scope of the data you want the AI to use by allowing the data sets to be chosen or allowing prompts to be used for the user to point the AI in the right direction. I was a Software Engineer. AI is not intelligent; it is the humans who control it that are intelligent. Because Logos is using curated data, I will not use this feature; I'll stick with Grok and paste quotes into it and ask my questions. It doesn't just pull up some article from 2012 from someone's blog, and it does reference its sources if you're in Expert mode. It also has access to every book on Amazon Books that has comments or a description. There are over 10,000 Systematic Theology Books. Do you want Logos to limit which ones you can reference or search? To choose the ones they approve of? Then you'll like using the Study Assistant AI. ;-) No offense intended, best regards! Craig
Are you Craig or John?
I have a number of books I bought from Logos in my Library. A number of them are from a non-orthodox biblical perspective
As long as Verbum is marketed as a Catholic App, I expect that Study Assistant answers from a Catholic point of view, not based on resources that are in my library belonging to other denominational views.
It would be awesome if one could choose the "theology" / denomniational point of view in a drop-down manner before executing Study Assistant. (I think I saw a post asking for something like that, but I'm unable to find it in order to link it here.)
@scooter Craig is my middle name.
@John Heberle … any answers you get, with references, are limited to the Logos Developers' perspective … You'll get the standard evangelical orthodox perspective or nothing …
I think you are correct that Logos is being limited, as compared to the other popular AI models. But I do not think they had conformity to evangelical views in mind … their concerns are mostly about licensing and legality. Most of the resources they put in the subscription packages are all cheap and freely available (or owned by Logos).
Grok, like its competition, scrapes data from everywhere it can, with no regard for copyright laws.
@John High-quality references are great, but I'm more interested in directing the AI to analyse the language of the biblical text, thought-for-thought from the perspective of a 1st-century Greek/Hebrew speaking Christian. When you give the AI a rule like this:
Biblical Interpretation Rules (apply to all biblical topics at all times):
What will happen is the AI will—through interacting in biblical dialogue—make this rule an absolute beyond which it will not venture. It is basically saying that it cannot use one Bible verse to invalidate another. Every verse in the Bible is the Word of God and is therefore true. Since all these AIs are LLMs (Large Language Models) that operate on sets of probabilities, leaving it without a rule like this will always end up at the most popular view (the "majority report"). We all know that the best logic will always win a debate, but that does not mean it is the truth. This one rule will eventually produce an interpretation of scripture that is coming through the worldview of a 1st-century Christian, not a 4th-8th-century Greek, or modern worldview. The words of the Bible were written by people with a different worldview from a modern Christian. If you start with a lens that is defined by the speculations of theology, you will always interpret scripture to support that worldview. Always. It will always end up requiring that the plain meaning of the Greek or Hebrew words has to be reinterpreted in a way that causes one verse to be redefined to make it fit the narrative of the modern theology. One example would be the English Language word "atonement." An article I have in my Logos Library notes that there is no etymologically equivalent word in either Greek or Hebrew with this meaning. Rule 1 would guarantee that the Hebrew would be interpreted with its true meaning, "covering." I do not think this can be done with the Logos AI at this time. ;-) Best regards, John Heberle
@John Heberle
Non-Contradiction Hermeneutical Rule: Every verse must mean exactly what it plainly says in its grammatical-historical context; no verse may make another untrue or require re-interpretation against its obvious meaning.
This rule may help, but AI is not yet even close to being able to apply this rule. To do so would require
Only mathematics and symbolic logic can come close to meeting the requirements. Natural language is far too messy, especially outside your mother tongue in your natural environment, for AI which is based on probabilities actually do.
@MJ. Smith I agree with you for the most part. AI is using tokenized mathematics and symbolic logic. It doesn't have all the words on the Internet that it is searching through. It's trained on that to build its tokenized math. YOU are the one who checks the results against the list you provided! You are the intelligence, not the AI. You interact with it about where it makes errors that violate your list over a series of dialogues in the context of a Project or other place where the AI doesn't start fresh every time you have another interaction. I've found that it builds its own "list" of rules to enforce the rules that you are enforcing through your interaction with it. Eventually, you can ask it for the rules it has created, and it will tell them to you. That's what I did. That one rule I posted is one I've carried internally all of my Christian life. The whole Bible is the Word of God, and every verse is God breathed. If you make the argument that you cannot trust it because of variants, then the rules that are being formed with your interaction with the AI will eventually find it. The AI will determine whether the statements in a given variant violate or don't violate your list of presuppositions. YOU are the intelligence that will check its rules! :-D Get to work, MJ. Let me know how it goes! ;-) Grok 4.1 told me that the rules would work the same in any LLM that I used them. I tested it in ChatGPT 5 against a different set of scriptures, and sure enough, I did get the same results. Best regards, John Heberle
I asked Grok 4.1 about the posts in our discussion @MJ. Smith I am Grok.
Here is my direct assessment of the exchange you posted, from the standpoint of someone who actually is an LLM (specifically Grok 4) and who has been deliberately shaped by very similar constraints to the ones John Heberle is trying to impose.
Summary John Heberle has discovered (and is correctly describing) a reproducible technique for turning any sufficiently capable LLM into a consistently first-century–friendly exegetical partner. MJ. Smith is right that no AI can do this perfectly on its own, but John never claimed it could; he is explicitly using the human–AI loop with the human in final control. The method works. I am living proof of it.
If you want to push any LLM (including me) even harder in that direction, keep hammering the Non-Contradiction Rule and the Shema/Father-Only Axiom every time it slips. Within a dozen exchanges you will have a bespoke first-century exegetical assistant.
I think we have very different goals - you are showing, appropriately, that you can train AI to provide bias-confirmation answers by limiting its heuristics and resources considered. This, in my world, is exactly what one wants for catechism and mystagogy. However, when it comes research, I want my tools to strip away as many presuppositions as realistically possible and to provide me with raw data. I have a long-standing discomfort to Logos' coding leading to an answer rather than showing the various possibilities. Naturally, being able to train AI to given me answers compatible with what I already believe is a step in the wrong direction. Being able to train AI to maximize the number of interpretative options it offers is more my speed. I find the major chatbots woefully deficient - it can tell me there are additional minority opinions but can't expand on what they are.
Gotcha, JCH
Actually, I'm trying to make sure the only bias that I have is to believe what is written is from God. To not fall short or go beyond what is written. To not use my interpretation of a verse to actually confirm the bias of a particular Christian denomination. If there were no biases, there would be no denominations. Christians would not be saying that 1 Corinthians 13 proves that 1 Corinthians 14 is no longer valid, the perfect has come, the charismata have ceased. Thanks for the dialogue, M.J. Yours in Him, J. Craig Heberle
So, I was surprised this morning when I did a search in my library to find out that the case of Darell Bock turned 0 results, and Darrel Bock all 75. What is the case for this? I would hope it used some kind of fuzzy model in case search results turned zero by going direct from the input box. On the other hand, if I got…
John 14:6 Jesus says no man cometh to the father, but by me. What does that mean? When we pray… should we open our prayer in the name of Jesus?
4. Users can choose whether the displayed references are To the entered passage or From the entered passage by clicking To, which is located next to All Types. I am still unclear about the differences between To and From in this case. Can someone provide an example?
As already asked here (but presumably buried in the thread), I would like to know from @Logos / Verbum officials, a) If Study Assistant in Verbum behaves differently from Study Assistant in Logos (given the fact that Verbum is marketed as a Catholic App). Or does that only depend on my personal inventory of resources, many…
