--- title: AIFL Interview Series: Geoffrey Stewart Morrison | daily.dev description: An interview with Geoffrey Stewart Morrison, professor at Aston University and Director of the Forensic Data Science Laboratory, traces his unconventional path... canonical: https://daily.dev/posts/aifl-interview-series-geoffrey-stewart-morrison-wsszrbkq3 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AIFL Interview Series: Geoffrey Stewart Morrison | daily.dev og:description: An interview with Geoffrey Stewart Morrison, professor at Aston University and Director of the Forensic Data Science Laboratory, traces his unconventional path... og:url: https://daily.dev/posts/aifl-interview-series-geoffrey-stewart-morrison-wsszrbkq3 og:image: https://api.daily.dev/og/posts/wssZrBkq3.png og:image:alt: AIFL Interview Series: Geoffrey Stewart Morrison og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # AIFL Interview Series: Geoffrey Stewart Morrison **[Medium](https://daily.dev/sources/medium_js)** · 13 min read · 0 upvotes · 0 comments ## Summary An interview with Geoffrey Stewart Morrison, professor at Aston University and Director of the Forensic Data Science Laboratory, traces his unconventional path from physics and theology through second-language speech perception research into forensic voice comparison. He discusses research showing forensic-voice-comparison systems outperform human listeners (individuals and groups) in speaker identification, his work on ISO 21043 forensic science standards, challenges in communicating likelihood ratios to judges and juries, and his advocacy for a 'forensic-data-science paradigm' that replaces subjective human judgment with calibrated, validated statistical and machine-learning models. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Do forensic-voice-comparison systems actually outperform human listeners at speaker identification? Yes, research comparing an automatic-speaker-recognition-based forensic-voice-comparison system against human listeners found none of the individual listeners performed as well as the system, even after being shown the system's output with an explanation. When groups of listeners conferred and made joint decisions, the system still outperformed roughly 80% of the groups. _Developers building evaluation benchmarks for ML systems against human baselines can find similar methodology debates on daily.dev._ ### What does it mean for a likelihood ratio system to be well calibrated in forensic evidence evaluation? A well-calibrated likelihood-ratio system produces output values with intrinsic meaning, analogous to a scale that reads 500 grams when 500 grams is placed on it. A poorly calibrated system might output a misleading number, like reading 30 for a 500-gram mass, making its results untrustworthy regardless of the underlying model's sophistication. _Those working on calibration and validation of statistical or ML models can explore related data science discussions on daily.dev._ ### Why do people tend to misjudge the strength of likelihood ratios presented as forensic evidence? Studies across multiple presentation formats and demographic groups found people consistently undervalue high likelihood ratios; a likelihood ratio of 1,000,000 was perceived on average as an effective ratio of less than 10. Participants appear to reweigh values based on contextual cues, such as believing DNA evidence is more trustworthy than shoe-mark evidence even when given identical numeric ratios. _Anyone studying how people interpret statistical confidence scores can follow related research on daily.dev._ ## Similar posts on daily.dev - [Fragments: March 26](https://daily.dev/posts/fragments-march-26-bby4yhjsi) · Martin Fowler · 1 upvotes · 0 comments - [4TB of voice samples stolen from 40,000 AI contractors -- how to verify if yours is being weaponized](https://daily.dev/posts/4tb-of-voice-samples-stolen-from-40-000-ai-contractors----how-to-verify-if-yours-is-being-weaponized-gv9bbkgcs) · Hacker News · 0 upvotes · 0 comments - [3 Questions: On the future of AI and the mathematical and physical sciences](https://daily.dev/posts/3-questions-on-the-future-of-ai-and-the-mathematical-and-physical-sciences-tvieufibv) · MIT News · 0 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning) [View this post on daily.dev](https://daily.dev/posts/aifl-interview-series-geoffrey-stewart-morrison-wsszrbkq3) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"AIFL Interview Series: Geoffrey Stewart Morrison","url":"https://daily.dev/posts/aifl-interview-series-geoffrey-stewart-morrison-wsszrbkq3","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/aifl-interview-series-geoffrey-stewart-morrison-wsszrbkq3"},"datePublished":"2026-10-04T10:07:22.732Z","dateModified":"2026-10-04T10:11:24.995Z","description":"An interview with Geoffrey Stewart Morrison, professor at Aston University and Director of the Forensic Data Science Laboratory, traces his unconventional path...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/d46320dd726351d701495bc1e604c1d5?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/d46320dd726351d701495bc1e604c1d5?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Medium","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Medium","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/medium","url":"https://daily.dev/sources/medium_js"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/aifl-interview-series-geoffrey-stewart-morrison-wsszrbkq3","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning","timeRequired":"PT13M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Medium","item":"https://daily.dev/sources/medium_js"},{"@type":"ListItem","position":3,"name":"AIFL Interview Series: Geoffrey Stewart Morrison"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/aifl-interview-series-geoffrey-stewart-morrison-wsszrbkq3#faq","mainEntity":[{"@type":"Question","name":"Do forensic-voice-comparison systems actually outperform human listeners at speaker identification?","acceptedAnswer":{"@type":"Answer","text":"Yes, research comparing an automatic-speaker-recognition-based forensic-voice-comparison system against human listeners found none of the individual listeners performed as well as the system, even after being shown the system's output with an explanation. When groups of listeners conferred and made joint decisions, the system still outperformed roughly 80% of the groups. Developers building evaluation benchmarks for ML systems against human baselines can find similar methodology debates on daily.dev."}},{"@type":"Question","name":"What does it mean for a likelihood ratio system to be well calibrated in forensic evidence evaluation?","acceptedAnswer":{"@type":"Answer","text":"A well-calibrated likelihood-ratio system produces output values with intrinsic meaning, analogous to a scale that reads 500 grams when 500 grams is placed on it. A poorly calibrated system might output a misleading number, like reading 30 for a 500-gram mass, making its results untrustworthy regardless of the underlying model's sophistication. Those working on calibration and validation of statistical or ML models can explore related data science discussions on daily.dev."}},{"@type":"Question","name":"Why do people tend to misjudge the strength of likelihood ratios presented as forensic evidence?","acceptedAnswer":{"@type":"Answer","text":"Studies across multiple presentation formats and demographic groups found people consistently undervalue high likelihood ratios; a likelihood ratio of 1,000,000 was perceived on average as an effective ratio of less than 10. Participants appear to reweigh values based on contextual cues, such as believing DNA evidence is more trustworthy than shoe-mark evidence even when given identical numeric ratios. Anyone studying how people interpret statistical confidence scores can follow related research on daily.dev."}}]} ```