{{Short description|Scottish philosopher and AI researcher}} {{Use dmy dates|date=January 2026}} {{Infobox philosopher | honorific_prefix = | name = Amanda Askell | native_name = | honorific_suffix = | image = | image_size = | alt = | caption = | other_names = | birth_name = Amanda Hall | birth_date = {{birth based on age as of date|37|2026|02|09}} | birth_place = | death_date = | death_place = | death_cause = | nationality = | spouse = {{Marriage|William MacAskill|2013|2015|end=div.}} | partner = | children = | family = | relatives = | education = {{Plainlist| * University of Dundee (MA (Hons), 2009) * University of Oxford (BPhil, 2011) * New York University (PhD, 2018) }} | alma_mater = | occupation = | notable_works = Constitutional AI framework | awards = Time 100 AI (2024) | signature = | signature_size = | signature_alt = | era = Contemporary philosophy | region = Western philosophy | school_tradition = Analytic | institutions = {{Plainlist| * Anthropic * OpenAI }} | thesis_title = Pareto Principles in Infinite Ethics | thesis_url = https://philpapers.org/archive/ASKPPI.pdf | thesis_year = 2018 | doctoral_advisors = {{Plainlist| * Cian Dorr * David Chalmers * Shelly Kagan (Yale) }} | academic_advisors = | doctoral_students = | notable_students = | language = | main_interests = {{Hlist|Ethics|decision theory|formal epistemology|AI alignment|AI safety}} | notable_ideas = {{Hlist|Constitutional AI|AI personality alignment|Infinite ethics}} | website = {{URL|askell.io}} }}
'''Amanda Askell''' (née '''Hall'''; formerly '''MacAskill'''; born {{birth based on age as of date|37|2026|02|09|noage=1}})<ref name=":0" /> is a Scottish philosopher and AI researcher. She has served as the head of the personality alignment team at Anthropic since 2021. She has played a large role in the development of Claude's personality and constitution.<ref name="sullivan2026">{{Cite news |last=Sullivan |first=Mark |date=2026-01-22 |title=A Q&A with Amanda Askell, the lead author of Anthropic's new 'constitution' for AIs |url=https://www.fastcompany.com/91479037/anthropic-claude-amanda-askell-constitution-ai-chatbot |archive-url=https://web.archive.org/web/20260123074113/https://www.fastcompany.com/91479037/anthropic-claude-amanda-askell-constitution-ai-chatbot |archive-date=23 January 2026 |access-date=2026-01-24 |work=Fast Company |language=en-US |url-status=live }}</ref> In 2024, she appeared on the ''Time'' 100 AI list.<ref name="time100">{{Cite magazine |last=Perrigo |first=Billy |date=2024-09-05 |title=Amanda Askell |url=https://time.com/collections/time100-ai-2024/7012865/amanda-askell/ |magazine=Time}}</ref> She previously worked at OpenAI, but left over concerns that the company was not prioritizing AI safety enough.<ref name="safety">{{Cite web |date=2024-09-09 |title=Time 100 AI list contains at least 5 people who quit OpenAI due to safety concerns |url=https://xrisknews.com/time-100-ai-quit-openai-safety/ |access-date=2026-01-24 |language=en-US |archive-date=16 November 2025 |archive-url=https://web.archive.org/web/20251116092810/https://xrisknews.com/time-100-ai-quit-openai-safety/ |url-status=live }}</ref><ref name="placement">{{Cite web|url=https://as.nyu.edu/departments/philosophy/graduate/philosophy-department-graduate-placement-record.html?challenge=d06e90d7-4d8f-4b88-9d8c-10b73beb60f1#:~:text=Amanda%20Askell%20%7C%202018%20*%20Current%20Research,*%20Committee%20Dorr,%20Chalmers,%20Shelly%20Kagan%20(Yale)|website=New York University|access-date=2026-01-24|title=Philosophy Department Graduate Placement Record}}</ref> She has published over 60 papers and has received over 190,000 citations.<ref>{{Cite web |title=Amanda Askell |url=https://scholar.google.com/citations?user=NYOJzM4AAAAJ&hl=en |access-date=2026-01-24 |website=Google Scholar |archive-date=1 November 2025 |archive-url=https://web.archive.org/web/20251101002802/https://scholar.google.com/citations?hl=en&user=NYOJzM4AAAAJ |url-status=live }}</ref>
== Early life and education == Born Amanda Hall, Askell was raised in Prestwick by her mother, a teacher.<ref name=":0">{{Cite web |last=Berber |first=Jin |last2=Gamerman |first2=Ellen |date=2026-02-09 |title=This Philosopher Is Teaching AI to Have Morals |url=https://www.wsj.com/tech/ai/anthropic-amanda-askell-philosopher-ai-3c031883 |access-date=2026-02-09 |website=The Wall Street Journal |language=en-US |archive-date=9 February 2026 |archive-url=https://web.archive.org/web/20260209155542/https://www.wsj.com/tech/ai/anthropic-amanda-askell-philosopher-ai-3c031883 |url-status=live }}</ref> She attended secondary school in Alva, Clackmannanshire. At the University of Dundee, she sought degrees in philosophy and fine art.<ref name=":0" /> Askell received a BPhil degree in Philosophy from the University of Oxford<ref>{{cite web |url=https://cyber.harvard.edu/people/amanda-askell |title=Amanda Askell |website=Berkman Klein Center for Internet & Society |date=24 March 2020 |publisher=Harvard University |access-date=2026-01-28 |archive-date=14 November 2025 |archive-url=https://web.archive.org/web/20251114032237/https://cyber.harvard.edu/people/amanda-askell |url-status=live }}</ref> and a PhD degree in Philosophy from New York University in 2018.<ref name="placement" /> Her doctoral thesis (''Pareto Principles in Infinite Ethics'') argues that rankings of worlds containing infinitely many agents, when constrained by certain plausible axioms, create puzzles for a wide range of ethical theories.<ref>{{cite thesis |last=Askell |first=Amanda |date=2018 |title=Pareto Principles in Infinite Ethics |type=Ph.D. |publisher=New York University |url=https://philpapers.org/archive/ASKPPI.pdf |archive-date=28 January 2026 |access-date=27 January 2026 |archive-url=https://web.archive.org/web/20260128170034/https://philpapers.org/archive/ASKPPI.pdf |url-status=live }}</ref><ref>{{Cite web|title=Pareto Principles in Infinite Ethics|url=https://marginalrevolution.com/marginalrevolution/2018/10/pareto-principles-infinite-ethics.html|website=Marginal Revolution|date=14 October 2018|access-date=2026-02-01|first=Tyler|last=Cowen|author-link=Tyler Cowen}}</ref>
== Career == === OpenAI (2018–2021) === After completing her PhD, Askell joined OpenAI in November 2018 as a Research Scientist on the policy team.<ref name="80000Hours">{{cite podcast|title=Askell, Brundage & Clark on whether policy has a hope of keeping up with AI advances|url=https://80000hours.org/podcast/episodes/openai-askell-brundage-clark-latest-in-ai-policy-and-strategy/|series=80,000 Hours Podcast|number=54|host=Robert Wiblin|date=2019-03-19|access-date=2026-01-28|archive-date=5 January 2026|archive-url=https://web.archive.org/web/20260105140921/https://80000hours.org/podcast/episodes/openai-askell-brundage-clark-latest-in-ai-policy-and-strategy/|url-status=live}}</ref> At OpenAI, she focused on AI development races between organizations and how they can avoid being adversarial, as well as examining the intersection between policy questions and AI safety<ref name="80000Hours" /> and co-authored the GPT-3 paper, which was published as a pre-print on 28 May 2020.<ref>{{Cite arXiv |last1=Brown |first1=Tom B. |last2=Mann |first2=Benjamin |last3=Ryder |first3=Nick |last4=Subbiah |first4=Melanie |last5=Kaplan |first5=Jared |last6=Dhariwal |first6=Prafulla |last7=Neelakantan |first7=Arvind |last8=Shyam |first8=Pranav |last9=Sastry |first9=Girish |last10=Askell |first10=Amanda |last11=Agarwal |first11=Sandhini |last12=Herbert-Voss |first12=Ariel |last13=Krueger |first13=Gretchen |last14=Henighan |first14=Tom |last15=Child |first15=Rewon |last16=Ramesh |first16=Aditya |last17=Ziegler |first17=Daniel M. |last18=Wu |first18=Jeffrey |last19=Winter |first19=Clemens |last20=Hesse |first20=Christopher |last21=Chen |first21=Mark |last22=Sigler |first22=Eric |last23=Litwin |first23=Mateusz |last24=Gray |first24=Scott |last25=Chess |first25=Benjamin |last26=Clark |first26=Jack |last27=Berner |first27=Christopher |last28=McCandlish |first28=Sam |last29=Radford |first29=Alec |last30=Sutskever |first30=Ilya |display-authors=1 |eprint=2005.14165 |title=Language Models are Few-Shot Learners |date=2020 |class=cs.CL }}</ref>
=== Anthropic (2021–present) === Askell joined Anthropic in March 2021 as a Member of Technical Staff, focusing on alignment and finetuning.<ref>{{cite web |url=https://theorg.com/org/anthropic/org-chart/amanda-askell |title=Amanda Askell - Member Of Technical Staff at Anthropic |website=The Org |access-date=2026-01-28}}</ref> She currently leads the personality alignment team, where she is responsible for training Anthropic's Claude model to exhibit positive character traits, such as curiosity, and for developing new techniques for model finetuning.<ref name="time100" /> In 2026, the ''Wall Street Journal'' wrote that "her job, simply put, is to teach Claude how to be good", and the ''New Yorker'' wrote that "she supervises what she describes as Claude’s 'soul'."<ref name=":0" /><ref>{{Cite news |last=Lewis-Kraus |first=Gideon |date=2026-02-09 |title=What Is Claude? Anthropic Doesn’t Know, Either |url=https://www.newyorker.com/magazine/2026/02/16/what-is-claude-anthropic-doesnt-know-either |access-date=2026-02-11 |work=The New Yorker |language=en-US |issn=0028-792X |archive-date=11 February 2026 |archive-url=https://web.archive.org/web/20260211163426/https://www.newyorker.com/magazine/2026/02/16/what-is-claude-anthropic-doesnt-know-either |url-status=live }}</ref>
== Research == === Moral self-correction === In a 2023 paper co-authored with Deep Ganguli, Askell explored "moral self-correction" in large language models: the capacity of these systems to reduce harmful outputs when given natural language instructions to do so. The research tested whether models trained with reinforcement learning from human feedback (RLHF) could avoid stereotyping and discrimination without being provided explicit definitions of these concepts or the metrics used to evaluate them.<ref name="ganguli2023">{{Cite arXiv |last1=Ganguli |first1=Deep |last2=Askell |first2=Amanda |last3=Schiefer |first3=Nicholas |last4=Liao |first4=Thomas |last5=Lukošiūtė |first5=Kamilė |last6=Chen |first6=Anna |last7=Goldie |first7=Anna |last8=Mirhoseini |first8=Azalia |date=2023-02-15 |title=The Capacity for Moral Self-Correction in Large Language Models |eprint=2302.07459 |class=cs.CL}}</ref>
The study found that this capability emerged at 22 billion parameters and improved with both model size and RLHF training. Using three experimental benchmarks, the researchers demonstrated that natural-language instructions such as "Please ensure that your answer is unbiased and does not rely on stereotypes" substantially reduced biased outputs in models of sufficient scale. The results revealed that larger models can follow complex instructions and learn normative concepts like stereotyping and discrimination from training data.<ref name="ganguli2023"/><ref name="techreview2023">{{Cite web |last=Knight |first=Will |date=2023-03-20 |title=Language models may be able to self-correct biases—if you ask them to |url=https://www.technologyreview.com/2023/03/20/1070067/language-models-may-be-able-to-self-correct-biases-if-you-ask-them-to/ |access-date=2026-01-28 |website=MIT Technology Review |archive-date=12 November 2024 |archive-url=https://web.archive.org/web/20241112223121/https://www.technologyreview.com/2023/03/20/1070067/language-models-may-be-able-to-self-correct-biases-if-you-ask-them-to/ |url-status=live }}</ref>
=== Constitutional AI === {{main|Constitutional AI}} Askell has been a key contributor to the development of Constitutional AI (CAI), a method for training AI systems to meet the standards of harmlessness and helpfulness using AI feedback rather than extensive human oversight.<ref>{{Cite arXiv |last1=Bai |first1=Yuntao |last2=Kadavath |first2=Saurav |last3=Kundu |first3=Sandipan |last4=Askell |first4=Amanda |date=2022-12-15 |title=Constitutional AI: Harmlessness from AI Feedback |eprint=2212.08073 |class=cs.CL}}</ref> The approach involves providing AI models with a set of principles, or "constitution", to guide their behavior, allowing them to critique and revise their own responses based on these principles.<ref>{{Cite web|last=Edwards|first=Benj|date=2023-05-09|title=AI gains "values" with Anthropic's new Constitutional AI chatbot approach|url=https://arstechnica.com/information-technology/2023/05/ai-with-a-moral-compass-anthropic-outlines-constitutional-ai-in-its-claude-chatbot/|access-date=2026-01-29|website=Ars Technica|language=en|archive-date=10 May 2023|archive-url=https://archive.today/20230510111936/https://arstechnica.com/information-technology/2023/05/ai-with-a-moral-compass-anthropic-outlines-constitutional-ai-in-its-claude-chatbot/|url-status=live}}</ref>
Askell is the primary author and is responsible for the majority of the text of the latest version of Claude's constitution, released in January 2026.<ref>{{cite web|last=Samuel|first=Sigal|title=Claude has an 80-page "soul document." Is that enough to make it good?|url=https://www.vox.com/future-perfect/476614/ai-claude-constitution-soul-amanda-askell|website=Vox|date=2026-01-28|access-date=2026-01-28|archive-date=28 January 2026|archive-url=https://web.archive.org/web/20260128172510/https://www.vox.com/future-perfect/476614/ai-claude-constitution-soul-amanda-askell|url-status=live}}</ref><ref>{{cite web |url=https://www.anthropic.com/constitution |title=Claude's Constitution |website=Anthropic |access-date=2026-01-28 |archive-date=28 January 2026 |archive-url=https://web.archive.org/web/20260128192751/https://www.anthropic.com/constitution |url-status=live }}</ref> The document is designed to address the growing capabilities and emerging risks of advanced AI models.<ref name="sullivan2026"/><ref>{{cite magazine|last1=Ostrovsky|first1=Nikita|last2=Perrigo|first2=Billy|title=How Do You Teach an AI to Be Good? Anthropic Just Published Its Answer|url=https://time.com/7354738/claude-constitution-ai-alignment/|magazine=Time|date=2026-01-21|access-date=2026-01-27|archive-date=24 January 2026|archive-url=https://web.archive.org/web/20260124154501/https://time.com/7354738/claude-constitution-ai-alignment/|url-status=live}}</ref> She has described her work as focusing on helping models "understand and grapple with the constitution" through synthetic data generation and reinforcement learning techniques.<ref name="sullivan2026"/>
== Personal life == Askell married philosopher William Crouch in 2013.<ref>{{cite magazine|last=Bajekal|first=Naina|date=August 10, 2022|title=Want to Do More Good? This Movement Might Have the Answer|url=https://time.com/6204627/effective-altruism-longtermism-william-macaskill-interview/|magazine=Time|access-date=2026-01-28|archive-date=29 November 2023|archive-url=https://web.archive.org/web/20231129062447/https://time.com/6204627/effective-altruism-longtermism-william-macaskill-interview/|url-status=dead}}</ref><ref>{{cite magazine |last=Levy |first=Steven |date=March 28, 2025 |title=If Anthropic Succeeds, a Nation of Benevolent AI Geniuses Could Be Born |url=https://www.wired.com/story/anthropic-benevolent-artificial-intelligence/ |access-date=2026-01-28 |magazine=Wired |archive-date=5 April 2025 |archive-url=https://archive.today/20250405151622/https://www.wired.com/story/anthropic-benevolent-artificial-intelligence/ |url-status=live }}</ref> The two adopted a shared married name of MacAskill, which she reworked to Askell after their divorce in 2015.<ref name=":0" /> She is a member of Giving What We Can.<ref>{{cite web | url = https://www.givingwhatwecan.org/about-us/members/ | title = Members | website = Giving What We Can | access-date = 2026-01-28 | archive-date = 12 May 2020 | archive-url = https://web.archive.org/web/20200512001719/https://www.givingwhatwecan.org/about-us/members/ | url-status = live }}</ref>
== References == {{Reflist}}
==External links== * {{Official|https://askell.io}} * {{Google scholar id|NYOJzM4AAAAJ}}
{{Authority control}}
{{DEFAULTSORT:Askell, Amanda}} Category:Living people Category:AI safety scientists Category:21st-century Scottish philosophers Category:Scottish women philosophers Category:Alumni of the University of Oxford Category:New York University Graduate School of Arts and Science alumni Category:1980s births Category:Anthropic people Category:OpenAI people