Confidently Wrong: AI, Uncertainty, and Open Source

Confidently Wrong: AI, Uncertainty, and Open Source

This is a special episode of WTBAI in which Kimberly sits down with her former colleague Derek Hanson to unpack what language research reveals about today’s AI systems, and together they consider where builders risk going wrong. Kimberly brings a corpus linguistics lens to large language models, reframing them as pattern-recognition systems trained on messy, biased “corpora” of the web. Her early insight was that AI is as powerful for feedback as it is for generation, and that this is an important distinction for education, ethics, and product design. Drawing from her EdTech startup (Moxie), she explains how embedding linguistic frameworks (e.g., Swales’ move-step analysis [https://en.wikipedia.org/wiki/CARS_model]) enabled structured feedback ... until frontier models caught up. The conversation then turns to open source and WordPress, where AI integration is accelerating across a massive ecosystem. Key themes: * Corpus vs. model: what LLMs are actually sampling * “Normalized overconfidence” and confidently wrong outputs * Why feedback > generation in many real-world use cases * Guardrails, prompt design, and early “agent-like” systems * Auditability gap: code transparency vs. output transparency * Bias sources: training data + human annotators * Missing voices: humanities, age diversity, non-developers * Friction as a feature: slowing down for rigor and care * A critical question for builders: how does your system handle uncertainty? The practical takeaway for builders is that before shipping AI features, ask whether your system surfaces or suppresses uncertainty, and whether a human could actually defend its outputs. Links: * Women Talk About AI: https://womentalkaboutai.com * Kimberly Pace Becker (LinkedIn): https://www.linkedin.com/kimberlypacebecker * “Stochastic Parrots” paper (Bender et al., 2021): https://dl.acm.org/doi/10.1145/3442188.3445922 Leave us a comment or a suggestion! [https://www.buzzsprout.com/2411501/fan_mail/new] Support the show [https://www.buzzsprout.com/2411501/support] Contact us: https://www.womentalkinboutai.com/

Motherhood and Higher Ed Burnout in an AI Moment

In this episode, Kimberly Becker and Professor Laura Dumin pull back the curtain on motherhood, higher ed burnout, and AI's effects on teaching. They talk pretty candidly about midcareer life with Laura sharing the reality of juggling three internal grants, release time, her kids' summer camp rush, and student needs and Kimberly tracing her own path out of Moxie, the AI feedback startup she co-founded with Jessica, and into a job completely outside academia after half a year of applications with zero interviews. Together, they discuss rising intolerance for institutional nonsense and why higher ed initiatives often feel like yet another layer of unpaid labor. Key themes: * 4–4 teaching loads and the myth of “just add research” * Being the primary earner: health insurance, risk, and career choices * Closing an edtech startup and facing a brutal job market * Midlife in academia: burnout, boundaries, and “less tolerance for everything” * Why many of us are choosing “good enough” over constant hustle Suggested links to include: * LinkedIn profiles for Kimberly [http://www.linkedin.com/kimberlypacebecker] and Laura [https://www.linkedin.com/in/laura-dumin157/] * Prior WTBAI episode about Moxie  [https://www.buzzsprout.com/2411501/episodes/17723850] Leave us a comment or a suggestion! [https://www.buzzsprout.com/2411501/fan_mail/new] Support the show [https://www.buzzsprout.com/2411501/support] Contact us: https://www.womentalkinboutai.com/

The Pope Joins the Chat that Women Were Already Having

When Pope Leo XIV released Magnifica Humanitas [https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html], his landmark encyclical on AI and human dignity, it lit up LinkedIn, Substacks, and newsfeeds worldwide. Kimberly read it the morning it dropped. Jessica, whose complicated relationship with religious institutions runs deep, read it anyway. And both of us had the same reaction as Abi Awomosu's [https://abiawomosu.substack.com/]: women have been saying this, uncited.  In this episode, we explore the encyclical's arguments, like: * technology is never neutral * unchecked growth impoverishes rather than enriches * treating limitations as defects is a category error, and  * concentrated technocratic power may be beyond the reach of regulation.  And we also name what's missing: the women, the scholars of color, and the critics who were making these exact arguments years before the Vatican caught up. We draw threads from the Pope's letter through late-stage capitalism, the bread-and-circus dynamics of the attention economy, and what Jolene Blais called AI's role as a "catabolic agent. [https://www.buzzsprout.com/2411501/episodes/18406320]" We talk about certainty language, the death of expertise, and why scientists are trained to live with uncertainty (and why that training is increasingly under attack). We end up, somehow, at microplastics, frugal hedonism, egg freezing, and communes. It's that kind of episode. In this episode: * What encyclicals are and why this one matters — even if you're not Catholic * The specific passages we highlighted and why they resonated * Abi Awomosu's [https://abiawomosu.substack.com] critique: women have been saying this, uncited — and her piece "Vatican Washing: Why All the Tech Broligarchs' Roads Now Lead to Rome" [https://abiawomosu.substack.com] * The "Who Said It First" problem and why it's more complicated than it looks * Posthumanism and transhumanism, and the Pope's sharp warning about treating some lives as less worthy * Data centers, extractive infrastructure, and colonial parallels * Why scientists hedge (and why that's a feature, not a bug) * Late-stage capitalism, the disintegration of community, and why collective action is harder when the technology driving us apart is the same technology we'd need to organize against * Frugal hedonism as a form of resistance * Pit & Peach: Kimberly's mom heads back to Mississippi (with a plan), and Jessica takes her first step toward freezing her eggs References & Links The encyclical: * Magnifica Humanitas — Full text, Vatican.va [https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html] * Why is Anthropic helping launch the Pope's encyclical? — National Catholic Reporter [https://www.ncronline.org/news/why-ai-company-anthropic-helping-launch-pope-leo-xivs-encyclical] (co-founder Chris Olah spoke at the Vatican presentation — yes, really) Scholarship & criticism: * Abi Awomosu, "How Not to Use AI" — Substack [https://abiawomosu.substack.com] * Bender, Gebru et al., "On the Dangers of Stochastic Parrots" (2021) [https://dl.acm.org/doi/10.1145/3442188.3445922] — the paper Timnit Gebru was fired from Google over; a preview of nearly every argument that followed * Tom Nichols, The Death of Expertise (2017) [https://www.amazon.com/Death-Expertise-Campaign-Established-Knowledge/dp/0190469412] Books: * Klara and the Sun — Kazuo Ishiguro [https://www.penguinrandomhouse.com/books/653825/klara-and-the-sun-a-gma-book-club-pick-by-kazuo-ishiguro/] * He, She and It — Marge Piercy [https://margepiercy.com/he-she-and-it] — feminist cyborg novel from 1991 that remains eerily prescient on AI, corporate power, and community * The Art of Frugal Hedonism — Annie Raser-Rowland & Adam Grubb [https://frugalhedonism.com/] From our archives: * WTBAI: "The Trojan Horse of AI" with Jolene Blais & Jon Ippolito [https://www.buzzsprout.com/2411501/episodes/18406320] * Our paper in Frontiers in Education: AI as Cultural Intermediary  [https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2026.1827603/full?sfnsn=scwspwa] Leave us a comment or a suggestion! [https://www.buzzsprout.com/2411501/fan_mail/new] Support the show [https://www.buzzsprout.com/2411501/support] Contact us: https://www.womentalkinboutai.com/

AI Voice Cloning: Trust, Persuasion, and Who's at Risk

When you call your bank, your doctor's office, or your financial planner, the voice that greets you may have been deliberately engineered to make you feel safe, calm, and compliant — and you almost certainly can't tell. Research shows [https://scholar.google.com/scholar?q=human+synthetic+voice+detection+accuracy] people correctly identify synthetic voice only about 55% of the time. That's barely better than a coin flip. In this co-host deep dive, Kimberly and Jessica pull apart what "voice" actually is (pitch, pace, prosody, timbre, accent) and why those features matter for trust, persuasion, and power. Synthetic voice isn't new, but the technology has crossed a threshold because it now replicates the subtle features that signal warmth, authority, and credibility. That has obvious applications in healthcare and customer service. It also powers grandparent scams, deepfake executive impersonation, and sales pipelines designed to move you from skepticism to compliance before you notice what happened.  In this episode: * What linguistics actually tells us about why we trust certain voices (and why politicians hire coaches to lower their pitch) * The FTC's 2024 numbers on imposter scams [https://www.ftc.gov/reports/consumer-sentinel-network] — $700 million lost by people over 60 in one year, a 362% increase from 2020 * The Hong Kong finance worker who wired ~$25 million USD (HK$200 million) after a deepfake CFO appeared on a Zoom call * ElevenLabs [https://elevenlabs.io/], Speechify [https://speechify.com/], and the companies building what they call "emotional operating systems" for AI * Trust vs. persuasion: when shared goals protect you — and when they don't * Why older adults are the highest-risk population, and why detection tools aren't the solution * Where regulation actually stands: New York's synthetic performer law (SB 7013) [https://www.nysenate.gov/legislation/bills/2023/S7013], the EU AI Act [https://artificialintelligenceact.eu/], and what's still missing * Practical questions to ask yourself — and the companies you interact with Mentioned in this episode: * Klara and the Sun [https://www.penguinrandomhouse.com/books/653825/klara-and-the-sun-by-kazuo-ishiguro/] by Kazuo Ishiguro * Project Hail Mary [https://www.imdb.com/title/tt12042730/] directed by Drew Goddard, starring Ryan Gosling (film, 2025) * The Martian [https://www.andyweirauthor.com/books/the-martian-hc] by Andy Weir * "Walk my Walk" by Blanco Brown (the real human artist) [https://youtu.be/EputaD-xMrY?si=IZjSOWUZT7sSDXIJ] * "Walk my Walk" by Breaking Rust (the AI-generated version) [https://youtu.be/OU71XDWYeIk?si=fC28Tle8X4k4W2ES] * Kimberly and Jessica's paper: "Defining and assessing AI literacy for researchers across the research lifecycle [https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2026.1827603/full]" in Frontiers in Education  Leave us a comment or a suggestion! [https://www.buzzsprout.com/2411501/fan_mail/new] Support the show [https://www.buzzsprout.com/2411501/support] Contact us: https://www.womentalkinboutai.com/

The Certainty Trap: Why the AI Future Isn't Already Written

In this episode, we sit down with Dr. Julia Stamm [https://www.linkedin.com/in/dr-julia-stamm/], founder and CEO of She Shapes AI [https://www.sheshapes.ai/], to unpack "The Certainty Trap." The way tech leaders project inevitability about AI, and the way that projection strips the rest of us of our agency. Julia is a sociologist and has held senior roles at the European Commission and the G20.  We talk about why so many AI adoption strategies are measuring the wrong things, why employees are quietly doing more work since AI showed up rather than less, and why women founders keep getting penalized for running for-profit businesses while their male counterparts get celebrated for the same thing. Julia also shares why she believes the most powerful question any of us can ask right now is simply, who benefits from this story being told this way? Topics Covered * The certainty trap and Julia's TEDx talk on reclaiming agency in the AI age [https://www.youtube.com/watch?v=E632scG95dY] * Why the inevitability narrative is marketing, not prophecy * The for-profit double standard that women founders face * How AI adoption is breaking the social fabric of organizations * Why measuring adoption rates and time saved are the wrong metrics * The magic triangle behind She Shapes AI: female leadership, responsible AI, and social impact * Real examples of women building AI for impact, including Rhiana Spring's Sophia chatbot [https://www.sheshapes.ai/2025/26-finalists] for survivors of domestic violence * Why employees are doing more work, not less, since AI arrived * The loss of optimism about the future and what it means for how we talk about AI * Why seeking out alternative narratives matters, and where to find them Referenced in This Episode * She Shapes AI [https://www.sheshapes.ai/] * Julia's TEDx talk: Beyond the Certainty Trap [https://www.youtube.com/watch?v=E632scG95dY] * She Shapes AI Global Awards 2025/26 finalists [https://www.sheshapes.ai/2025/26-finalists] * Rest of World [https://restofworld.org/], the nonprofit publication covering technology stories beyond the West * Empire of AI [https://www.penguinrandomhouse.com/books/743569/empire-of-ai-by-karen-hao/] by Karen Hao * Cory Doctorow on the TINA framework (there is no alternative) * Ethan Mollick on the 3% of organizations using AI in the sweet spot * Julia Stamm on LinkedIn [https://www.linkedin.com/in/dr-julia-stamm/] * Julia Stamm on Substack [https://substack.com/@juliastamm] * Julia's forthcoming personal website at juliastamm.com  Leave us a comment or a suggestion! Leave us a comment or a suggestion! [https://www.buzzsprout.com/2411501/fan_mail/new] Support the show [https://www.buzzsprout.com/2411501/support] Contact us: https://www.womentalkinboutai.com/