Operationalizing Pesak, part vi: Gemir (and Sevir) as Necessary and Sufficient
This is an ongoing series trying to understand what is required for valid pesak halacha, in order to explore if ChatGPT or some other AI can pasken; and if so, what sorts of capabilities must be present.
In the first post in this series, I mentioned gemir and sevir as required qualities for pesak halacha. This qualities appear in different contexts, for instance for judges in monetary cases. And these qualities might be different for specific contexts, so the gemir of three laymen judging might be different than the gemir of an ordained judge in terms of knowledge of halacha.
We’re not actually concerned overmuch, for now, with judging monetary cases. In terms of ruling practical ritual law — can I perform this action on Shabbat; is this chicken kosher; how do I construct this sukkah; essentially, what they would ordain someone for in yoreh yoreh semicha, there is a requirement of gemir and sevir.
As I noted in that post, this is found in Sanhedrin 5b,
יוֹרֶה, יוֹרֶה. אִי גְּמִיר, רְשׁוּתָא לְמָה לִי לְמִישְׁקַל? מִשּׁוּם מַעֲשֶׂה שֶׁהָיָה.
With regard to the permission granted to Rabba bar Ḥana and Rav, the Gemara had related that Rabbi Ḥiyya asked Rabbi Yehuda HaNasi: May he teach people and issue rulings concerning what is prohibited and what is permitted? And Rabbi Yehuda HaNasi responded: He may teach. The Gemara asks: If he had studied and mastered the relevant halakhot, why do I need him to receive permission? The need for formal authority is understandable when it comes to serving on a court to judge cases of monetary law, but any knowledgeable person should be qualified to answer questions about ritual law. The Gemara explains: The need for such permission is due to an incident that took place.
דְּתַנְיָא: פַּעַם אַחַת הָלַךְ רַבִּי לְמָקוֹם אֶחָד, וְרָאָה בְּנֵי אָדָם שֶׁמְּגַבְּלִין עִיסּוֹתֵיהֶם בְּטוּמְאָה. אָמַר לָהֶם: מִפְּנֵי מָה אַתֶּם מְגַבְּלִין עִיסּוֹתֵיכֶם בְּטוּמְאָה? אָמְרוּ לוֹ: תַּלְמִיד אֶחָד בָּא לְכָאן וְהוֹרָה לָנוּ: מֵי בְצָעִים אֵין מַכְשִׁירִין. וְהוּא מֵי בֵיצִים דְּרַשׁ לְהוּ, וְאִינְהוּ סְבוּר מֵי בְצָעִים קָאָמַר.
As it is taught in a baraita: Rabbi Yehuda HaNasi once went to a certain place, and he saw people there kneading dough while they were in a state of ritual impurity, and they believed that nevertheless, the dough remained ritually pure. Rabbi Yehuda HaNasi said to them: For what reason are you kneading your dough in a state of ritual impurity? They said to him: A certain Torah scholar came here and taught us that water from swamps [mei betza’im] does not render food susceptible to contract ritual impurity. Therefore, they would take water from swamps and knead dough with it, in the mistaken belief that such dough would not be susceptible to ritual impurity. But in reality, what he taught them was that water of eggs [mei beitzim], i.e., the albumin of eggs, does not render food susceptible to impurity, as it is not considered water. But they thought he said: Water from swamps.
From this gemara, it seems like “gemir” is necessary, and might be defined as knowledge of the halacha, or knowledge of the relevant sources.
This sugya does not explicitly mention sevir here for hora’a, but it might go along with it, given that we saw gemir and sevir accompany one another in Rav Nachman’s statement about his ability to judge monetary cases alone. However, we also pointed to Horayot 2b, where for a moment, the idea of severing gemir from sevir was entertained, in terms of a student fit for ruling halacha, תַּלְמִיד וְרָאוּי לְהוֹרָאָה, who could have objected to the Sanhedrin’s erroneous ruling. Rava seems to allow gemir but not sevir as a valid entity who could have objected, as well as a sevir but not gemir; Abaye objects that הוֹרָאָה implies both.
Sevir might be defined as a level of competent reasoning ability — for instance, being able to apply knowledge of sources, and of underlying halachic principles, to a particular new case.
Different depth of knowledge and reasoning might be required for halachic questions of different sorts, as we’ll eventually get to. To be able to answer a question which is a simple lookup — “can I cook this egg which has been laid on Yom Tov? I forgot Yaaleh Veyavo in Shacharit Shmoneh Esrei, so do I need to repeat?” might require a relatively low level of gemir. Applying to common situations which are akin to discussed ones might require a different level — “which items are similar to an egg”. Applying to new realities — “is an LED light OK for havdalah?” or balancing competing halachic values, e.g. “I forgot Yaaleh Veyavo, but it is Maariv which may have status of reshut, and I am the Shaliach Tzibbur who the people will have to wait for, and I’ll be embarrassed” might require deeper knowledge or reasoning abilities.
We haven’t concretely defined gemir and sevir, but let us just run with the plausible definitions above.
The point I wanted to make in this post is that, from these sugyot in Sanhedrin and Horayot, it seems that:
(1) Gemir and sevir are necessary qualities, without which one should not pasken.
(2) Gemir and sevir are sufficient qualities, and we don’t truly need something like ordination, some official appointment to the job of paskening.
What suggests point 2 is the gemara’s question of רְשׁוּתָא לְמָה לִי לְמִישְׁקַל if he is גְּמִיר. The gemara’s answer validates the question — that such official ordination isn’t strictly necessary, but it was a measure put in place to ensure a different critical quality, ability to clearly and unambiguously communicate.
We could try to put this into a cheftza vs. gavra framework. It is not that legitimate pesak arrives from a valid gavta, a person. And that only an ordained “posek” can produce a halachically significant entity, a ruling called a pesak.
Rather, a pesak halacha is just a correct halachic answer. We are results oriented, and if the correct answer can be reasoned out / calculated / known, and we can be confident that it is correct, then we can rely on it. It does not matter what human or other entity produced it, or what process produced it.
I think that this is what Rabbi Herschel Schachter meant when he discussed the legitimacy of an AI paskening in the Headlines podcast, as I summarized in my Torah To-Go article, and that this gemara or something similar is prompting it. Assuming that we can be confident that the AI — be it an expert system, a reasoning system, a database lookup, or perhaps even a LLM — produces or calculates accurate results, it should not matter that it is not a human. Pesak is just the right answer; it isn’t something that has significance only if produced by a halachically valid gavra.
(Of course, we’d need to grapple with gemaras in Horayot about what constitutes a halachically significant though inaccurate hora’ah of beit din, such that an individual who relied on their erroneous ruling can be encompassed in their korban. But Rav Schachter certainly has weighed in on Horayot sugyas, and surely has a way of explaining that — for instance, in terms of the laws of korbanot and the reasonableness in toleh beda’at acheirim. So too other ideas Rav Schachter has discussed of multiple halachically valid systems of pesak that can coexist, as in the Sanhedrin of each shevet, or Ashkenazim vs. Sefardim, because variety is the spice of life, and this is one form of eilu ve’eilu. That might suggest that there isn’t one unique correct answer, and all we need is to find it. This is all resolvable, but we aren’t going to get into it here.)
By a similar token, and lehavdil, I think I recall heard Rav Schachter speak about women paskening, for instance as a Yo’etzet Halacha, and him making a similar narrow point in the context of a greater discussion. But a woman who knows and can reason can pasken in this sense; even a non-Jew who knows and can reason can pasken in this sense. (Even though he raises be other concerns, such as breadth of knowledge, knowing what other aspects of halacha are relevant, which aren’t found in particular sections of Shulchan Aruch they have mastered.) To put into our terms, it is not the gavra, but the cheftza, or the result. Bli neder, I hope to link to and discuss in the future.
In contrast, Rabbi Netanel Wiederblank has an article in that same YU Torah To-Go, Ten Reasons Why We Cannot Turn to AI for Psak: Understanding the Nature and Philosophy of the Halachic Process, where he writes:
Psak is not a preexisting thing. It is created by the posek. It is not the answer to the question; it is a human answer to the question given be a qualified posek, someone higiya l-hora’ah.
This is a gavra-oriented approach. And Rabbi Wiederblank doesn’t just assert it — he backs it up with sources, such as tannur shel achnai, which he interprets in particular ways, and which I’ll put in my queue to eventually discuss. How could we, and how would Rav Schachter, deal with those sources? I think I already know, and alluded to it above.