Law as Translation: Courts, Practitioners, and the Limits of AI

Introduction

A court reads a treaty as a contract. A lawyer’s silence is mistaken for incompetence. An innovative software provides a verisimilar yet incorrect legal answer. While variant in context, these scenarios share a point of failure. Law fails when translation fails; courts misinterpret treaties across legal traditions, lawyers lose clients, and technologies underdeliver when they ignore that law, especially as an increasingly interjurisdictional practice, is not a neutral set of rules governing conduct but a practice of carrying meaning across systems. While anthropologists, such as Clifford Geertz, and legal scholars including Lawrence M. Friedman, Grant M. Hayden, and George Mousourakis, have long argued that law must be reframed as cultural system and understood within its cultural context, mainstream education, doctrine, and practice have failed to shed the impulses of efficiency and profit [1]. This blind spot matters: when institutions, professionals, and technologies across the legal landscape overlook the importance of culture in legal reasoning, they risk incoherence in judgment and incompetence in practice.

This essay argues that law is best understood as cultural translation, with culture constituting, for the purposes of this essay, the shared and socially transmitted behaviors, norms, and values of a given group that operate independently of individual preference [2]. Culture’s translation in turn transcends simple linguistic substitution and comprises a process of understanding and addressing each layer of culture underlying behavior. The essay utilizes the U.S. Supreme Court’s decision in BG Group v. Argentina (2014), a first-of-its-kind investor-state arbitration case, to demonstrate how judges reframe treaty clauses across legal traditions, altering meaning in the interpretative process. A set of oral histories from cross-border and cross-cultural lawyers illuminates how the task of translation permeates each stratum of legal systems, requiring that each practitioner, from a student to a justice, adapt their voice, style, and framework to successfully carry law across rifts. The rise of artificial intelligence (AI) finally proves this point in reverse: machines can process texts at superhuman speed, relegating menial processes of the legal profession to be historical relics. However, their failure to reduce the legal tapestries of cultural meaning, woven from millennia of lived narratives, cultural codes, and human relationships, into algorithms capable of feeling and deploying translation, exposes the law’s uniquely human core. These perspectives reveal that translation is not peripheral to law, but rather, it is law. It is the fundamental process through which judges, lawyers, and AI models succeed or fail in translating treaties across doctrinal frameworks, themselves across professional cultures, or reasoning across human boundaries respectively.

As professionals incorporate AI into legal work and education, judicial decision-making, they must ask not only what efficiencies are gained and what costs are cut, but also what nuances of judgment may be compromised in the process. As with its past, the future of law cannot be captured in one language, system, or algorithm. It will be written, instead, in translation. The success of future systems depends on whether professionals identify and preserve translation as the irreducibly human core of law.

Courts as Translators: BG Group v. Argentina

Few cases highlight the law’s translational nature more obviously than BG Group v. Argentina (2014), in which the U.S. Supreme Court sought to determine whether an arbitral tribunal had exceeded its powers by bypassing a limited-term litigation requirement of the investment treaty between the U.K. and Argentina [3]. The case is widely regarded as the Supreme Court’s first major encounter with investor–state arbitration, a field that itself is hybrid: balancing state sovereignty with private ordering, and public treaties with contractual logics [4]. Scholars of arbitration have long emphasized that this hybridity is central to the field’s very nature, shaping it as a transnational legal order [5]. The pertinent reflection of this case lies not in the precedent set for future arbitration clauses, but rather in the Justices’ obscuring of the cultural stakes in their decision and presenting their readings as neutral doctrine, when, in fact, they were actively conveying meaning across cultural and systemic boundaries. Their opinions reveal the extent of this translation.

Among its articles establishing procedure for resolving contract disputes, Article 8 of the 1990 U.K.-Argentina Bilateral Investment Treaty (BIT) provided that disputes must first be submitted to Argentine courts; only in the absence of a final decision within eighteen months could the investor then initiate arbitration [6]. In the present case, BG Group skipped this delineated “waiting period,” arguing that such a process would be unduly slow and expensive [7]. The arbitral tribunal agreed, triggering a review in U.S. courts which ultimately culminated at the Supreme Court [8]. Although each step in this process provides an individual case study into the different interpretations of the same clause, the split decision of the Supreme Court best captures the discourse writ large.

In the Supreme Court’s reversing the Court of Appeals’ decision, Justice Breyer, author of the majority opinion, framed the question as one of timing rather than consent [9]. Legal scholar Alan Rau classifies this opinion as a transformation of the language of public international law into a procedural precondition familiar to common-law arbitral practice [10]. His view that arbitrators had the primary authority to interpret the requirement thus constitutes a fundamental transposition. Professor Anthea Roberts and Barrister Christina Trahanas further sharpen this critique by noting that the court’s decision analogized a bilateral international treaty, under which investors were not direct signatories but rather third-party beneficiaries, to a commercial contract, thereby malforming public law and its foundations to support the decision doctrinally [11]. From yet another vantage point, Professor Andrea Bjorklund of McGill University underscored the uneasy fit between national courts and international instruments, highlighting how a domestic decision can reshape a treaty’s global meaning [12]. The majority opinion ultimately made common law presumptions to bring a public law treaty into a more comfortable realm.

In contrast, Chief Justice Roberts’ dissent, as joined by Justice Kennedy, drew attention to the absence of consent to arbitration between BG Group and Argentina until the eighteen-month dispute resolution precondition was satisfied [13]. Consent of the state to arbitration could only be presumed after the investor complied with the treaty’s local litigation requirement. To treat Article 8’s condition as an avoidable hurdle rather than a requirement was thus to conjure consent to arbitration where none existed [14]. The dissent, in contrast to the majority opinion, performed a different kind of translation; it reframed the BIT as a rigid pact oriented by state sovereignty and public international law. As with the former interpretation, however, Chief Justice Roberts’ approach was not neutral but instead pursued through a different cultural and, perhaps less obviously, pragmatic compass. This reading, while preserving the treaty’s indigenous spirit, risked rendering the condition as an impenetrable shield for the host state to wield at will.

Justice Sotomayor’s concurrence charted a middle path between the majority and dissent [15]. Although her opinion aligned with Justice Breyer’s conclusion, given that the opinion did not impact the future interpretation of such treaties, she nevertheless raised the concern of how the efficacy of such “conditions of consent” could be safeguarded. Her analysis sought to balance practical implications on commercial interests with abstract discussions of sovereignty. In this way, the concurrence echoes Political scientist Alec Stone Sweet’s account of arbitration’s “judicialization [16].” Arbitration is ever evolving. As it increasingly permeates contracts as the most common form of dispute resolution, professionals at each stage of the process must hone their practice through concepts of fairness, efficiency, and utility. Justice Sotomayor’s concurrence reflects this exact trend in its translating Article 8 with simultaneous regards for the material end produced by the interpretation and the proportionality of the means through which that is accomplished.

Across these opinions, unique approaches to translating across systemic and cultural boundaries create disparities in issues highlighted and conclusions reached. Justice Breyer translates the treaty directly into the context of U.S. arbitration law. This act of domestication demonstrates how courts, in the process of translation, risk subordinating the transnational character of investment arbitration, rooted in treaty-based consent and arbitral autonomy, to the comfort of domestic judicial presumptions and arbitral doctrines. Arbitration expert Gary Born, from a doctrinal standpoint, read the majority’s approach as persuasive precisely because it carried arbitral logic into the treaty context [17]. Chief Justice Roberts translates it into a larger question of sovereignty transcending the dispute between the present parties. Justice Sotomayor translates the agreement into a pseudo-pragmatic application. Each application reflects different cultural assumptions and systemic orientations, and BG Group v. Argentina thus exposes the need for translation-conscious adjudication. The difference between the opinions stems not from doctrinal disagreements but fundamentally different characterizations of the agreement’s nature. While this does not inherently implicate the Court’s decision to be unreasonable, the counterfactual of a translation-conscious exemplifies a scenario in which frameworks of understanding could be chosen and debated more openly, and the spirit of international systems could be carried through with less readjustment. Ultimately, the process and result of the case serve not as an indictment of subpar conduct but rather as an exemplar for how applying the framework of law as translation bolsters analytical scenarios in which facts, interests, and agreement transcend a singular language, jurisdiction, or culture. This lesson implicates not just the highest level of decision-making but permeates down to the fundamental building blocks of legal practice: students and early professionals.

Professionals as Translators: From the Classroom to the Courtroom

If BG Group v. Argentina exposes how courts translate transnational legal texts, the experiences of cross-border practitioners proves that lawyers translate themselves. The interviews conducted for the Bridging Barriers project highlight, in conjunction with present legal anthropological discourse, that cross-border practice is not mechanical transfer but translation [18]. This section argues that lawyers are not neutral technicians who transfer knowledge from one system to another, but cultural translators who must continually adapt voice, style, and conduct across divergent legal registers. Recognizing this translational work is critical not only for understanding practice but also for rethinking how legal education prepares students for transnational careers.

When juxtaposed through individual transitions, Turkiye’s civil and the United States’ common law systems illustrate manifestations of translation from the classroom to the courtroom. Three Turkish legal practitioners, Sezgi Guler, Bukem Gebelek, and Murat Ersin, reflect on the prevalence of translation in their experiences. Guler’s reflections highlight the most fundamental translation from a civil-law to a common-law system [19]. Turkish legal education, steeped in the tradition of various civil codes and regarded with the veneration resultant, had conditioned her to view the classroom as a space of reverence. In Istanbul, this expectation of respect manifested in measured silence; in Boston, the same silence signified indolence or incompetence. To be seen as capable and extract benefit from her schooling, she had to shed her training and adopt an intellectually adversarial posture. By learning different contextually expected methods of engagement, she developed the capacity to navigate across systems. Her experience reflects what professor George Mousourakis stresses: comparative law is about recognizing law as a cultural phenomenon, embedded in traditions and values [20]. As scholars Zweigert and Kötz remind us, law cannot be confined to “rules on the books” but must be studied within the customs and practices that give those rules meaning [21].

Bukem Gebelek’s experience underscores the manifestation of expectations in writing [22]. Turkish pedagogy, in her experience, rewarded the exhaustive utilization of codes and the ornamentation of exposition. In combination, she found these demands of writing to be implicit proofs of one’s education and competence. U.S. legal writing demanded something different: concision and persuasion. While both systems value analytical rigor, their approaches to the role of writing in professional processes differed significantly. Gebelek’s ability to discern these differences and adapt to the requests of professors and clients constitute an exemplar of the translation necessary to individual traversal across systems. Her experience confirms what professors Shiyuan Han and Reinhard Zimmermann describe: divergences in private law are not merely doctrinal but cultural, shaped by deep-seated traditions regarding what constitutes legal authority [23].

This need for adaptation transcends the written word to every aspect of practice, as evidenced by Murat Ersin’s focus on different presentations of professionalism [24]. Turkish practice emphasizes formality in dress, hierarchy, and manner of address, while American practice tolerates and even expects informality as a mark of confidence. His navigation of these competing standards exemplifies professor Pierre Legrand’s claim that legal meaning and procedure are so deeply rooted in culture that they resist straightforward translation [25]. At the same time, his ability to recalibrate suggests what professor Ralf Michaels defends: that functional comparative analysis can still identify workable patterns across systems without erasing their cultural specificity [26].

These contrasts between Turkish civil-law and American common-law education underscore that translation is not only linguistic but structural. The same professional acts, such as speaking in class, drafting legal analysis, and presenting oneself to colleagues, acquire different meanings depending on the cultural context, constituting a labor of translation together. Lawyers are not passive transmitters of legal knowledge, but rather active agents who reconstruct meaning across various systems. Law itself must thus be understood within its cultural and historical context [27]. Despite these expectations, doctrinal mastery and efficiency continue to dominate contemporary legal education. If courts must adopt translation-conscious adjudication, law schools must also adopt translation-conscious pedagogy, explicitly cultivating the comparative and adaptive skills that lawyers already practice informally. To prepare lawyers for a global profession is to train them as translators.

AI as a Translator: Aid over Arbiter

If courts translate texts and lawyers translate identities, AI reveals what happens when translation is sidelined in favor of efficiency. The arrival of AI in law has been celebrated with predictions of reviews being completed, contracts being drafted, and correspondences being sent all within minutes. Its promise of speed, efficiency, and cost saving have been well received, as according to the Thomson Reuters institute 77%, 74%, and 58% of lawyers used AI to review, summarize, and compose documents and contracts, respectively, in 2025, in sharp increases from a year prior [28]. Market competition will bring these numbers closer to ubiquity in the next few years as professionals fear becoming sidelined. The algorithms and repositories of these models will thus revolutionize the throughput of the profession writ large. Yet, as the preceding sections have demonstrated, law is not reducible to neutral words to be processed. It is a human practice of cultural translation. Although dazzling, the capabilities that make AI so revolutionary—pattern recognition, speed, and fluency—mask its core deficiency in its inability to live and feel across the linguistic, cultural, and systemic zones of legal practice. Reliance on AI thus underscores the centrality of human expertise even as machines process languages at scale [29].

Legal education and practice are already testing AI’s capacities. ChatGPT has been shown to produce passing-level answers on real law school exams [30]. More recently, randomized controlled trials have demonstrated that while GPT-4 assistance only slightly and inconsistently improves the quality of legal analysis, it produces significant gains in speed and student satisfaction among law students [31]. Scholars reinforce this point from different angles. On the one hand, Frank Pasquale’s The Black Box Society (2015) offers an early warning: algorithmic systems are opaque “black boxes,” ill-suited for domains—like law—that demand transparency, accountability, and cultural sensitivity [32]. The latter of these demands stands out especially as even though AI can, through inputs outlining context, capture and emulate certain behaviors, this mimicking is limited to the extent that culture can be reduced to verbal descriptions. The lived component of culture thus eludes AI. Harry Surden concludes in parallel that while AI can process vast datasets with speed, it falters on the cultural and value-driven dimensions of law [33]. Most recently, Lee F. Peoples confirms these concerns empirically, finding that large language models can imitate the structure of legal reasoning but remain shallow, inconsistent, and prone to error [34]. Taken together, these incapabilities are what distinguish AI tools from seasoned professionals. Although AI can reliably produce vast quantities of documents far exceeding human counterparts, its failure to polish doctrinal competence with cultural calibration precludes it from successfully completing tasks that require cultural adaptation. For example, AI-generated memos may undermine the relationship between cooperating parties because they ignore different cultural norms of hierarchy and deference or fail to persuade a common-law audience by relying on bulky and systemic civil-law analysis; such cases demonstrate how legally accurate outputs that lack cultural adaptation may still fail as legal communication. Ultimately, without refinement of models or reinforcement by humans, AI simply renders legal work into structural mimicry, stripping it of judgment, analysis, and understanding.

Similar challenges in other professions allow for a comparative understanding. As Miranda and colleagues observe, medical algorithms are able to outperform doctors in every arena of knowledge and do so with a far greater efficiency [35]. This disparity is already by orders of magnitude, and the cost is pennies on the dollar [36]. Yet, as Dhruv Khullar emphasizes, such systems are a place to start, not to end [37]. Their results change depending on what questions are asked, revealing that doctors remain indispensable, not merely to interpret outputs, but to frame the inquiry itself. This contingency makes clear that diagnostic power alone cannot substitute for judgment. And even a lay observer can identify that humans can provide what machines by their nature cannot: empathy, reassurance, and communication. The communication of a technical diagnosis to the patient requires a translation into realistic optimism and lived meaning, similar reassurances that, although less obvious in law, allow for the effective collaboration of letter-of-the-law lawyers and business partners. Education, too, underscores this. Howard Gardner argues that by 2050, AI will displace three of his five fundamental aspects of mind, the disciplined, synthesizing, and creating, and leave behind two minds that parallel the insufficiency of a diagnostic AI, the respectful and ethical minds [38]. Algorithms can and will master the processing-based cognitive processes but cannot and will not usurp the human capacities of carrying meaning across differences.

The proper role of AI in law, as tracked through medicine and education, must then be an aid of efficiency rather than an arbiter of meaning. Like personal computers as a revolutionary addition to office spaces, AI can be invaluable in streamlining processing, lowering costs, optimizing cognitive efficiency, while also serving as a guiding force in legal education and an accelerator of doctrinal development. But it must be a tool of human processes; efficiency without translation is not law; it is verbal computation.

Conclusion

Across courtrooms, classrooms, and offices, one lesson emerges: law is undeniably an act of translation. BG Group v. Argentina demonstrates that even at the highest levels of judicial debate, discourse can shatter into competing interpretations because of underlying differences in the understanding of pertinent concepts. Interviews with younger, cross-border legal professionals highlight that this trend is not confined to higher levels of the legal hierarchy. The increasing prevalence of careers across languages, cultures, and jurisdictions requires that their lawyers translate themselves and their experiences across each of their interactions. AI demonstrates the converse in its limitations. Lacking the deeply human qualities of practice, AI collapses law into a faux efficiency bound by words on paper to be processed.

These disparate experiences also hint to a set of common implications. Courts must adopt adjudication that is explicitly framed through translation as they increasingly handle greater quantities of cases lacking comfortable assumptions of domestic systems. Law schools must cultivate pedagogies that intentionally prepare students to not only master doctrine and reasoning but also navigate cultural differences with agility and fluency, equipping them for the profession of tomorrow. Practitioners for their part must safeguard their role as translators, developing a symbiotic, rather than displacing, relationship with a fast-developing array of AI tools. To endure and excel amid unprecedented globalization and automation, legal institutions and professionals must recognize and adopt translation not as a tangent but as the irreducible core of law. For law lives only in translation: the unending work of carrying meaning across systems with care, humility, and judgment.

Footnotes

[1] Clifford Geertz, Local Knowledge: Further Essays in Interpretive Anthropology (New York: Basic Books, 1983), chap. 8, “Fact and Law in Comparative Perspective,” 167ff.; Lawrence M. Friedman and Grant M. Hayden, American Law: An Introduction, 3rd ed. (New York: Oxford University Press, 2017), 6 and 123 ff.; George Mousourakis, Comparative Law and Legal Traditions: Historical and Contemporary Perspectives, 2nd ed. (London: Routledge, 2020), chap. 6, sec. 3, 137ff.

[2] Luis Cardenas, “Overcoming Cultural Barriers in Negotiations and the Importance of Communication in International Business Deals,” Program on Negotation, January 8, 2026, https://www.pon.harvard.edu/daily/international-negotiation-daily/bridging-the-cultural-divide-in-international-business-negotiations/.

[3] BG Group PLC v. Republic of Argentina, 572 U.S. 25 (2014) [also reported at 134 S. Ct. 1198].

[4] Nigel Blackaby et al., Redfern & Hunter on International Arbitration, Student Version, 6th ed. (Oxford: Oxford University Press, 2015), chap. 1, pp. 1–27, chap. 7, pp. 415–439, chap. 8, pp. 441–446.

[5] Yves Dezalay and Bryant G. Garth, Dealing in Virtue: International Commercial Arbitration and the Construction of a Transnational Legal Order (Chicago: University of Chicago Press, 1996), chap. 1, pp. 3–17.

[6] Agreement Between the Government of the United Kingdom of Great Britain and Northern Ireland and the Government of the Republic of Argentina for the Promotion and Protection of Investments, Dec. 11, 1990, 1765 U.N.T.S. 33.

[7] Anthea Roberts and Christina Trahanas, “Judicial Review of Investment Treaty Awards: BG Group v. Argentina,” American Journal of International Law 108, no. 4 (October 2014): 750–63.

[8] Ibid.

[9] 572 U.S. 25 (2014) at 33-35 (Breyer, J., majority opinion).

[10] Alan Scott Rau, “BG Group and ‘Conditions’ to Arbitral Jurisdiction,” Pepperdine Law Review 43 (2016): 577–614.

[11] Anthea Roberts and Christina Trahanas, 754.

[12] Andrea K. Bjorklund, “BG Group and ‘Conditions’ to Arbitral Jurisdiction,” Pepperdine Law Review 43 (2016): 615–36.

[13] 572 U.S. 25 (2014) at 45 and 61 (Roberts, C.J., dissenting).

[14] Anthea Roberts and Christina Trahanas, 754-55.

[15] 572 U.S. 25 (2014) at 50 (Sotomayor, J., concurring).

[16] Alec Stone Sweet, “Arbitration and Judicialization,” Oñati Socio-Legal Series 1, no. 9 (2011): 1–21, https://ssrn.com/abstract=1988923.

[17] Gary B. Born, International Commercial Arbitration, vol. 1 (Alphen aan den Rijn: Kluwer Law International, 2nd ed. 2014), 1003–1006.

[18] Can Yildirim, Bridging Barriers: Conversations Across Legal Cultures, https://www.aycaakkayanyildirim.com/en/bridging-barriers-conversations-across-legal-cultures/

[19] Can Yıldırım, “Bridging Barriers No. 1– from Istanbul to Boston: What Sezgi Guler Taught Me about Speaking up and Thinking across Borders,” Ayça Akkayan Yıldırım, July 22, 2025, https://www.aycaakkayanyildirim.com/en/bridging-barriers-no-1-from-istanbul-to-boston-what-sezgi-guler-taught-me-about-speaking-up-and-thinking-across-borders/.

[20] Mousourakis, Comparative Law and Legal Traditions, 139ff.

[21] Konrad Zweigert and Hein Kötz, An Introduction to Comparative Law, 3rd rev. ed., trans. Tony Weir (Oxford: Oxford University Press, 1998), 36–39.

[22] Can Yıldırım, “Bridging Barriers No. 2– Introducing Bukem Gebelek: Building a Global Legal Career across Systems,” Ayça Akkayan Yıldırım, August 5, 2025, https://www.aycaakkayanyildirim.com/en/bridging-barriers-no-2-introducing-bukem-gebelek-building-a-global-legal-career-across-systems/.

[23] Forewords of Shiyuan Han and Reinhard Zimmermann in James Gordley, Hao Jiang, and Arthur Taylor von Mehren, An Introduction to the Comparative Study of Private Law (Cambridge: Cambridge University Press, 2021), xx–xxii and xxiii-xxv.

[24] Can Yıldırım, “Bridging Barriers No. 4– Takeaways from My Conversation with Murat Ersin,” Ayça Akkayan Yıldırım, September 9, 2025, https://www.aycaakkayanyildirim.com/en/bridging-barriers-no-4-takeaways-from-my-conversation-with-murat-ersin/.

[25] Pierre Legrand, “The Same and the Different,” in Comparative Legal Studies: Traditions and Transitions, ed. Pierre Legrand and Roderick Munday (Cambridge: Cambridge University Press, 2003), 240–311.

[26] Ralf Michaels, “The Functional Method of Comparative Law,” in The Oxford Handbook of Comparative Law, ed. Mathias Reimann and Reinhard Zimmermann, 2nd ed. (Oxford: Oxford University Press, 2019), 345–75, at 349–50.

[27] Friedman and Hayden, American Law, 6; Mousourakis, Comparative Law and Legal Traditions, 137ff.

[28] “2025 Generative AI in Professional Services Report,” Thomson Reuters, April 7, 2025, https://www.thomsonreuters.com/content/dam/ewp-m/documents/thomsonreuters/en/pdf/reports/2025-generative-ai-in-professional-services-report-tr5433489-rgb.pdf.

[29] Daniel Schwarcz and Jonathan H. Choi, “AI Tools for Lawyers: A Practical Guide,” 108 Minnesota Law Review Headnotes 1 (2023), https://papers.ssrn.com/sol3/papers.cfm?abstract\_id=4404017.

[30] Jonathan H. Cho et al., “ChatGPT Goes to Law School,” Journal of Legal Education 71, no. 3 (Spring 2022): 387-400.

[31] Jonathan H. Choi, Amy B. Monahan, and Daniel Schwarcz, “Lawyering in the Age of Artificial Intelligence,” Minnesota Law Review 109 (2024): 147–218.

[32] Frank Pasquale, The Black Box Society: The Secret Algorithms That Control Money and Information (Cambridge, MA: Harvard University Press, 2015).

[33] Harry Surden, “Artificial Intelligence and Law: An Overview,” Georgia State University Law Review 35, no. 4 (2019): 1305–1337.

[34] Lee F. Peoples, “Artificial Intelligence and Legal Analysis: Implications for Legal Education and the Profession,” arXiv preprint (February 2025), https://arxiv.org/abs/2502.03487.

[35] J. Miranda et al., “Artificial Intelligence Outperforms Physicians in General Medical Knowledge, Except in the Paediatrics Domain: A Cross-Sectional Study,” Bioengineering (Basel) 12, no. 6 (June 14, 2025): 653, https://doi.org/10.3390/bioengineering12060653.

[36] Lauren Martin et al., “Better Call GPT: Comparing Large Language Models Against Lawyers,” arXiv preprint (2024), https://doi.org/10.48550/arXiv.2401.16212.

[37] Dhruv Khullar, “If A.I. Can Diagnose Patients, What Are Doctors For?,” The New Yorker, September 29, 2025, https://www.newyorker.com/magazine/2025/09/29/if-ai-can-diagnose-patients-what-are-doctors-for.

[38] Sy Boles, “How Ai Could Radically Change Schools by 2050,” Harvard Gazette, September 19, 2025, https://news.harvard.edu/gazette/story/2025/09/how-ai-could-radically-change-schools-by-2050/.