The quick development of AI will present the question in the near future: can we rely on artificial intelligence to make halachic rulings (psak halacha)?
In order to answer this question, we must first define what our intentions are when using the term psak halacha. And indeed, this tern has two primary meanings. First is the one included in halachic literature such as the Rambam and Shulchan Aruch. There we determine “laws” and the normal equations to follow. The second interpretation is the one included in responsa literature, which provides halachic responses and practical guidelines to follow in reality. For this literature, the halachic authority applies the “laws” determined by halachic literature unto a certain reality. Rulings such as these require analysis of a specific case and its possible interpretations as well as general knowledge in the realm of halachic norms so that it will be possible to properly tailor the approach to the specific case in light of halachic principles. Understandably, this differentiation is not clear cut, for halachic works adopt principles that are written in responsa literature and even deal with specific cases, and vice versa – responsa determine general principles as well.
The theoretical possibility of using AI continues to grow in both of these domains. In other words, both in determining new principles by combining knowledge in prior sources while adding the creative power of a halachic authority and in determining which halachic principles are relevant to the current case. Granted that there is still a long journey before AI will be able to do both of these actions properly in a way that will be indifferentiable from a great rabbi’s responsa, however it is speedily advancing towards this reality.
The halachic position could reason that halacha is not related just to the content, but also has an “institutional” factor, according to which only a skilled rabbi can rule halacha. If this position is accepted by Torah-observant Jews, the various tools of artificial intelligence can act as a information bank that can help study rabbinic rulings or analyze their content, however their rulings will have no halachic significance. According to this approach, halachic ruling is not only the process of making conclusions based on sources, but its application by a halachic authority. The authority can use the capabilities of AI in order to find sources to discuss the issue, compare his position to the position of other rabbis and learn what the generally accepted position is in the matter, but at the end of the day the halachic ruling will come from him – from his analytic mind, his spiritual world, and his interpretation of the sources and reality that he himself assesses.
What is the “Proper” Halachic Opinion?
This position, which I believe is the only one accepted by all of the rabbinic world nowadays, can encounter two essential struggles. First is the need to be accepted by the general public. Rabbi Abraham Isaac Kook defined the public’s status in halachic ruling – “the accepted norm of the public” in his words – as a central foundation at its core. Thus, from a practical perspective, the forefront of halacha is rooted in the opinions of halachic authorities, from Moshe Rabbeinu to the rabbinic greats of our generation, but if the public does not abide by this ruling – it will remain on the bookshelf or in the computer files, with no tangible influence.
This reality creates a dynamic and even problematic situation, where there is a divide between that which the authorities have determined and that which is done in practice, however one cannot deny that this happens often. This gap is even expressed in various mishnayot, where it is determined, for example, that matters are performed “against the wishes of the Sages”. The halachic analysis is the source of the Sage’s authority, but the practical implementation is the source of the public’s strength.
Second is the claim about the core issue: machine learning AI acts from a tremendous data supply of halachic rulings from rabbis in the past, and even in the halachic domain there are works that collect and summarize, functionally acting as an anthology and not an innovative creation. Why then should AI have a lesser portion, for it does the same thing, and utilizes rulings made by rabbis already established according to the “institutional” criterion? Is the only claim preventing AI from joining the halachic authorities of the next generation that a rabbi hasn’t made the ruling, or is the claim also in relation to its content?
It is difficult to answer this question, primarily because we do not know how to estimate the future technological advancements and how much it will truly be impossible to differentiate with precise care between rulings of man and machine. This matter first and foremost requires determining what the “proper” halachic ruling is, for if there is no criterion such as this – it will not be possible to discuss the issue of whether a machine is capable of ruling “correctly”. How can we determine if it interprets reality “correctly” and if its halachic guidance is “correct” when we ourselves oftentimes disagree and AI is just another opinion?
Even the claim that halachic ruling is not an action based on an algorithm but the experience of an authority with worlds of emotion, internalization, education, outlook to the future, world views, etc., and therefore one must reject AI which is unable to do this – is not clear cut for two reasons: first, as stated, it is possible that AI will be able to incorporate these factors into halachic rulings, in light of the millions of halachic responsa written in the past. Even if AI itself does not “feel”, it can learn how emotions work and even utilize them. As stated, we are dealing with a situation where we cannot assess where the “emotion” is coming from – whether from the world of the authority or from the databank. Beyond this, even in the halachic world there are more “formal” authorities, that almost do not incorporate these factors into their ruling at all.
Searching for an Answer
Despite this, I tend to oscillate between hoping and estimating that in the end, the God-fearing will not truly turn to artificial intelligence. This emerges from two domains that we have mentioned – the institutional domain and the content itself– however it also emerges from the conclusion we derive from the first story of halachic ruling – Parshat Yitro. Moshe Rabbeinu explains to Yitro that one who comes for judgement does not do so because they need a practical solution to a situation they are in, but in order “to seek God”. In other words: when confronting a rabbinic authority, even if this is done in writing or through responsa given over virtually, and all the more so in physical meetings, there is an additional soul beyond the practical domain. It is spiritual connection with another person.
This person, the halachic authority who has merited to be a part of the chain of halachic authorities throughout the generations, makes contact with the seeker and touches upon their character, life experience, and personal world. Even a written responsum in a book includes a relationship beyond the practical question under discussion. This connection between people brings an additional value with it which touches upon the soul. Even if one cannot identify it and determine that this responsum was written by man or AI, the knowledge that it was written by a person, and all the more so if written personally from a rabbi to a specific question asker – carries with it the additional soul of “like water is one face to another”.
This seemingly is the primary reason why artificial intelligence can provide additional tools for a rabbi to rule halacha, to challenge him, to propose its “position”, etc., however this relationship will forever be a basis of connected souls together, for the service of God and the keeping of his Torah.
