Can I speak lashon hara to ChatGPT?
On the surface, the question seems to have an easy answer: as impressive as AI is, it is not another person; one is not actually sharing the derogatory information with someone else. So you’d think the answer should certainly be yes.
However, there are a number of factors to consider.
One is that lashon hara is not only an offense of harm against another, but is considered a corruption of character as well. The Talmud identifies the speaker himself as one of three who are “killed” by lashon hara, along with the subject and the recipient (Arachin 15b). All sin is a source of harm to the sinner himself, in that through such actions the perpetrator incurs guilt and punishment. However, the implication here is that speaking lashi hara is unique in its impact on the actor, imposing an effect distinct from that of sin in general.
From a philosophical perspective, many Jewish thinkers have focused on the unique role of speech as a defining element of humanity. Speech, at least in its fully realized form, distinguishes Man from the animal; for humans, speech expresses thought, making this distinction especially profound. This notion is reflected in the Aramaic translation of Onkeles to the verse “and Man became a living soul” (Gen. 2:7), where the last words are rendered as “ruach mimalela” (“a talking spirit”). The Chafetz Chaim, in his introduction, cites this phrasing in explaining why he subtitled the main section of his work Mekor Chaim, or “Source of Life”.
Additionally, speech is an essentially nonphysical act with an immense impact on the physical world, and thus wields an awesome power (See Netivot Olam, netiv halashon, ch. 2, and Shemirat HaLashon, sha’ar hazechirah, ch. 1); “Death and life are in the hands of the tongue” goes the verse in Mishlei (18:21). Accordingly, the attribute of speech demands extreme care and sensitivity in its usage, and the abuse of this precious resource imposes an exceptional degradation on the speaker.
Further, on a level that is more psychological/moral, it has been asserted that offenses of speech both reflect and deepen particularly unsavory elements of the personality. Significantly, the speaker of lashon hara displays a bias toward negativity, a dangerous attitude which feeds upon itself and ultimately affects all that is in its purview. Maimonides writes as much (Mishneh Torah, Hil. Tuma’at Tzara’at 16:10), charting a direct path from hearsay to heresy, noting that those who at first are merely cynical in their orientation will eventually come to disparage and deny G-d Himself. The tendency toward a negative assessment is profoundly corrosive and, left unchecked, can undermine the possibility of a fair treatment and appreciation for any value of life.
Lashon hara represents at least two categories of evil: an act of damage, which is the offense against the subjects of the speech; and a base character trait, which is an aspect of the personality dangerous to the soul. (The psychological makeup of the personality inclined toward gossip is discussed in Rumor and Gossip: The Social Psychology of Hearsay (1976 by Ralph L. Rosnow and Gary Alan Fine), pp. 81-83).
Maimonides is often linked to the view that, in their legal formulation, mitzvot are defined in terms of actions – even if their ultimate purpose is to shape one’s thoughts or emotions. And that is why mitzvot usually command us to do or not do something, and not simply command a state of being. Nonetheless, he does speak of character traits in his legal code, in which he notes that a Torah scholar “judges his fellow favorably, speaks in praise of his friend and never to his disparagement” (Hil. Deiot 5:7. See Emek HaLashon, pp. 6-7). Many authorities, including the Chafetz Chaim (Hilchot Lashon Hara, klal 3, in Be’er Mayim Chaim #7), see this dual theme in the writings of Maimonides as well.
The character component may also be read into the format of the Torah’s core prohibition in this area. The impression created by a statement such as the one in Parshat Kedoshim: “Do not travel as a peddler” (Vayikra 19:16), rather than a more straightforward formulation such as “Do not disparage… ,” is that it is not an act that is being condemned, but rather a personality type; as if the instruction is not what not to do, but what not to be.
A further observation can be made from the fact that the Torah characterizes the act of gossip as “traveling as a peddler.” The Chafetz Chaim (Hil Lashon Hara, klal 1, BMC #4 and klal 2:12 in a footnote, citing the Shlah) offers the suggestion that the intent is to include not only the gossip itself but the prior acts leading up to it in the transgression. If so, the message would seem to again be that lashon hara is damaging not only because of the effect on the subject, but that the entire process and mentality are corrosive to the speaker.
All of this indicates that the very formation of the disparaging statements in one’s mind, apart from their communication to another human being, inflicts its own cost. This can also be seen from another aspect of the lashon hara prohibition.
Not only is there a prohibition against spreading lashon hara, but also “accepting” it (kabbalah) is also forbidden. According to the Talmud (Pesachim 118a; see also Mishneh Torah, Hil. Sanhedrin 21:7), this prohibition is located earlier in the Torah (Ex. 23:1), in a verse that is complicated to translate but appears to prohibit the “bearing” (lo tisa) of a report (sheima) that is “shav,” translated either as false or as unnecessary. In that same Talmudic passage, it is suggested that this verse addresses the speaker of lashon hara as well, as the report is “borne” by both the receiver and the transmitter.
The focus on personality corrosion independent of harm gives rise to a number of other issues. For example, lashon hara would be prohibited even if one could envision a situation in which the speaker could credibly maintain that the subject will not be harmed (Hil. Lashon Hara klal 3, # 6. See also Tehor Sefatayim, pp. 140-141, who understands the issue to be a dispute between Maimonides and Rabbenu Yonah. Note Resp. Az Nidberu, XIV, 59, who assumes that this prohibition would not include incidental discussion of information already known to the listener. Other than that category, however, he maintains in a later responsum (loc. cit. 65) that there is no lashon hara that can actually be characterized reliably as harmless).
Interestingly, while the issue of speaking lashon hara to “things” may be a newer question, the issue of speaking about things was discussed earlier, as may emerge from the condemnation of the spies and their slander of the Holy Land. From one perspective, this may stem from the association certain objects have with specific human beings, particularly if the object in question is merchandise used in business. However, this concern may also emerge from the corrosive effects of negativity and cynicism, even without a human target.
Further, there is an additional prohibition against netirah, or bearing a grudge. According to some understandings, one who acts to preserve animosity within his own heart runs afoul of this provision, leading some to suggest that merely keeping a journal in which such feelings are recorded could be problematic. Similarly, a conversation of this nature with AI would seem to raise similar issues.
Most significantly, it is not guaranteed that the information imparted to the AI will not cause any harm to the subject. This data does become absorbed by the system, and can be reflected back later in unhelpful ways to the original speaker, and possibly even to others, with untold consequences.
This last point is discussed at length in Chapter three – called “Artificial Gossips” – of the book Raising AI by computer scientist and ethicist De Kai. In the following passage, he uses the term “quidnunc” for the “hearing” party to gossip, what we would call a mekabel lashon hara.
As he writes, “The danger of artificial gossip is that it mushrooms regardless of whether it is true or false. When gossip is false or private, artificial gossip mongers spread fake news and confidential information, like human gossip mongers do, but exponentially more dangerously.… In this age of botnets, artificial gossips easily account for more than 25% of our gossips. Artificial gossips push past the tipping point for fake news to change society’s views. Just like humans, artificial gossip mongers weaponize the power of suggestion. Just like human gossips, they spread unevaluated claims that are based on superficial appearance. They promote hearsay rather than evidence, they promote stereotyping… Even when gossip is true, artificial quidnuncs gain social power just like human quidnuncs do, but again far more perilously…
“Have you joined the artificial gossips? As a parent, what kind of example are you setting for artificial children? Are you sharing gossip into the ears of artificial quidnuncs who use it to figure out how better to manipulate our society by propagating your gossip to other susceptible members of society? Are you avidly consuming gossip offered by artificial gossip mongers who thus encourage learning from you how to manipulate you to propagate more gossip that is even more enticing to you? Are you an unwitting part of the network? Have you joined the artificial gossips in creating divides rather than bridging them? Gossip ostracizes persons or groups. Exponential artificial gossip disruptively ostracizes persons or groups… You used to read news stories. Now the news stories read you. So don’t help artificial gossips. What they tell you might seem like simple clickbait, but it isn’t. It’s creating, over time, a system that will eat itself. Don’t be an unwitting part of the divisiveness.”
Due to all of the above, it cannot be simply maintained that speaking lashon hara to AI is a non-issue. That being said, lashon hara even to people is subject at times to justification, known as to’elet, or “purpose,” and such rationales would be relevant here as well, and in some cases would indicate that AI is a preferable choice as a recipient.
To’elet may include not only the thwarting of a predator, but also necessary benefit to the speaker. In this vein, it is likely that to relate negative information in the process of confiding in a therapist, or other helpful individuals, is justified. This point is noted by the Chafetz Chaim (Hil. Lashon Hara, 10:14 in a footnote, and see also 6:4) and actually draws upon two distinct forms of benefit.
The crucial realm of mental health is a self-evident priority that itself can justify many otherwise discouraged behaviors. More specifically, the benefit of talk therapy is identified in the Talmud (Yoma 75a), which in interpreting a verse (Prov. 12:25), advises, “One who finds worry in his heart should discuss the matter with others.” According to Rashi, the benefit is that the listener may have some solution to the problem. Within that interpretation, the purpose of such conversation is functional in the practical sense.
However, it is also likely that the intent is that the very act of talking is productive as an emotional support. Accordingly, such unburdening of the mind should be permitted even if the listener is not likely to offer concrete advice. It seems that this justification is grounded not only in the productive benefit but also in that there is no intent to disparage the individual being discussed, but rather to provide therapeutic relief to the speaker (See R. Zevulun Shuv, Sha’arei Zevulun, YD, 76).
The listener would not necessarily have to be a professional therapist if indeed the simple act of talking is deemed beneficial. Theoretically, such license could be extended to “venting,” if that is productive. It should be noted, however, that there is some debate among experts in psychology as to the actual benefit of venting anger (see Forgiveness Is a Choice, by Robert D. Enright, pp. 54-55; and see You Are Not So Smart, ch. 32, and the studies cited therein). Some studies indicate that “venting” either keeps initial anger running longer or causes an emotional dependency on the venting that otherwise would not be there.
R. Ya’akov Kamenetsky (Emet L’Ya’akov to Orach Chaim, 156, n182) distinguishes between one who is sharing his distress that an offender is going unpunished, which is permitted, and one who is using the publicizing of a misdeed as an expression of anger itself, which he maintains is prohibited. He compares such speech to the destructive behavior of one who smashes vessels in anger, implying that the speech is unproductive (or counterproductive) and thus not subject to a license of to’elet. (Compare Chelkat Binyamin, Hilchot Lashon Hara 10:39 and Netiv Chaim, Hilchot Lashon Hara 1:12, with Zera Chaim p. 308).
Centuries earlier, R. Yehudah HaChasid (Sefer Chasidim # 64) formulated this potential benefit, emphasizing also the perspective of the listener. He described a situation in which an angry individual is on the verbal warpath, determined to share his fury with the world. In such a case, while listening to lashon hara is normally discouraged, it may make sense for one to decide that he will choose to provide an audience for this individual, under the assumption that by doing so, he can reduce the speaker’s need to tell anyone else, while also playing a role in calming the speaker and trying to shift his perspective more positively.
The goal of reducing the number of listeners is an important one. A “venting license” should not be taken as a free pass to widely disparage the source of one’s anger. Emotional unburdening should be accomplished with a very limited number of people (preferably, one person) and should not involve the widespread dissemination of negativity (See Responsa LeChafetz BaChaim, I, 2, and II, 1).
In considering these particular benefits, ChatGPT may have some advantages as the “recipient” of the lashon hara. A helpful comparison may come from another frequent issue in the laws of lashon hara, the question of omitting names
It is popularly assumed that it is permissible to relay derogatory information as long as no names are used. However, some halachic authorities wondered if this was indeed the case, and some, in fact, concluded that such conversation was forbidden (see, for example, R. Akiva Eiger-Sofer, Responsa Hitorerut Teshuvah, Vol. I-II, 270. See R. Shabtai Sofer, Sha’arei Deah, to Mishneh Torah, Hil. Deiot, ch. 7, who understands this to be the position of Maimonides), while others did allow if it is clear there will be no harm caused (see R. Chaim Kanievsky, in She’ailat Rav I, ch 7 # 9). (It should be emphasized that the discussion here is relevant only to a situation where the identity is actually successfully concealed; if it is at all possible the identity would be perceived, all would agree that lashon hara is violated.)
Those who are stringent in this situation are apparently assuming that the concealing of the identities to protect the guilty does not eliminate the prohibited nature of the conversation. Indeed, just as a gossip column might relay salacious information without specifying the subject, merely stating that “a well-known politician” has been implicated in some scandal, there is prurient interest in such stories even without knowing the identities of the protagonists. This interest can nurture the negative character traits associated with lashon hara, and thus pose a problem even when damage is not being inflicted upon the subject (see also R. Aharon Roth, in the journal Marpei Lashon, II, pp. 16-19).
R. Shmuel Hominer, the author of a summary adaptation of the Chafetz Chaim’s writings, asserted (Ikarei Dinim, klal 3, in footnote) that relating lashon hara without the name of the subject may not be an act of “talebearing,” but is nonetheless a violation of character, and by reinforcing the perception of the subject in the mind of the speaker one commits an act of “acceptance” of lashon hara. He maintained that such gossip is addressed by the scriptural verses, “Let none of you plot evil against his brother in your heart” and “And let none of you plot evil in your hearts against his neighbor” (Zach. 7:10 and 8:17).
However, such conversation may be permitted according to all views if the intent is not to gossip but rather to convey an educational message or cautionary exhortation. In that case, as there is neither harm to the subjects nor a salacious motive, both elements of lashon hara appear to be mitigated. R. Hominer acknowledges this as well, but warns that it is necessary to ascertain that all listeners are aware of the motivation, so as to prevent any misunderstanding that would result in the perception of endorsing gossip.
Our ChatGPT question seems similar. There are certainly many factors that caution against a casual attitude of permissibility in this case. However, when there are justifiable goals to be accomplished, it may be argued that technology has provided a new avenue by which to minimize the problems and maximize the benefits. If so, this is one more development for which we can be grateful.