home.social

#bias — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #bias, aggregated by home.social.

fetched live
  1. I've been working on my #homelab and a question bothers me :thaenkin: :

    How can we integrate AI without reproducing or amplifying model biases ?

    We often say "garbage in, garbage out... (GIGO, 2020)"
    :blobpeek:
    But is it that simple ?

    LLMs reflect their training data, leading their inevitable biases.
    Does the model itself have a responsibility or is it 100% to the user ?

    I'm digging into the subject. How do you manage this in your projects ?

    #LLM #Bias #ethicalai #noai

  2. I've been working on my #homelab and a question bothers me :thaenkin: :

    How can we integrate AI without reproducing or amplifying model biases ?

    We often say "garbage in, garbage out... (GIGO, 2020)"
    :blobpeek:
    But is it that simple ?

    LLMs reflect their training data, leading their inevitable biases.
    Does the model itself have a responsibility or is it 100% to the user ?

    I'm digging into the subject. How do you manage this in your projects ?

    #LLM #Bias #ethicalai #noai #dev

  3. #Universities are moving away from #AIdetection tools due to concerns about #accuracy, #falsepositives, and #bias. While some institutions are redesigning assessments to emphasise student process and oral components, others still rely on detection. The lack of consistent policies and transparency around AI detection tools fuels student anxiety and raises questions about the role of surveillance in academic integrity. ft.com/content/49304b1e-8a9d-4 #AIagent #AI #ML #NLP #LLM #GenAI

  4. #Universities are moving away from #AIdetection tools due to concerns about #accuracy, #falsepositives, and #bias. While some institutions are redesigning assessments to emphasise student process and oral components, others still rely on detection. The lack of consistent policies and transparency around AI detection tools fuels student anxiety and raises questions about the role of surveillance in academic integrity. ft.com/content/49304b1e-8a9d-4 #AIagent #AI #ML #NLP #LLM #GenAI

  5. Accepting a hypothesis depends on the importance of being mistaken. It can be considered a judgment with ethics attached.
    Without accounting for multiple values and preferences in policy design, scientific insights risk becoming politicized, potentially reinforcing dominant parties’ interests in funneling action while shifting risks to vulnerable populations.
    Roger A. Pielke Jr. advocated for taking the role of saying "How about these other ideas?" instead of taking a position on currently debated topics.

    ref. (2007). "The Honest Broker: Making Sense of Science in Policy and Politics" 🧩 🧵

    #policy #uncertainty #uncertainties #robustness #probabilities #futures #anticipation #IAMs #science #modelling #modeling #risks #riskAssessment #unknowns #governance #bias #workCollectives #institutions #Pielke

  6. Accepting a hypothesis depends on the importance of being mistaken. It can be considered a judgment with ethics attached.
    Without accounting for multiple values and preferences in policy design, scientific insights risk becoming politicized, potentially reinforcing dominant parties’ interests in funneling action while shifting risks to vulnerable populations.
    Roger A. Pielke Jr. advocated for taking the role of saying "How about these other ideas?" instead of taking a position on currently debated topics.

    ref. (2007). "The Honest Broker: Making Sense of Science in Policy and Politics" 🧩 🧵

    #policy #uncertainty #uncertainties #robustness #probabilities #futures #anticipation #IAMs #science #modelling #modeling #risks #riskAssessment #unknowns #governance #bias #workCollectives #institutions #Pielke

  7. Paul Rudnick: "MAGA lackey Bari Weiss attended the White Correspondents Dinner where Trump called her 'a wonderful woman', after which she announced that from now on the CBS Evening News will be called 'Pres. Trump Tonight'."
    #media #CBSnews #bias.

  8. Paul Rudnick: "MAGA lackey Bari Weiss attended the White Correspondents Dinner where Trump called her 'a wonderful woman', after which she announced that from now on the CBS Evening News will be called 'Pres. Trump Tonight'."
    #media #CBSnews #bias.

  9. "The biggest danger of the Paramount-Warner mega-merger: The merger would deliver control of a second national TV news network to an owner with brazenly political ties. Paramount already owns CBS News and if it buys Warner, would own CNN as well":
    sfchronicle.com/opinion/openfo
    #MonopolizingSourcesOfNews #extremism #bias #politics #media

  10. "The biggest danger of the Paramount-Warner mega-merger: The merger would deliver control of a second national TV news network to an owner with brazenly political ties. Paramount already owns CBS News and if it buys Warner, would own CNN as well":
    sfchronicle.com/opinion/openfo
    #MonopolizingSourcesOfNews #extremism #bias #politics #media

  11. Calling out bias is one of the most powerful allyship tools, but it's important to know when and how to do it to avoid negative side effects and improve effectiveness. #wellness #bias #allyship

    psychologytoday.com/us/blog/be

    Posted into FLIPBOARD EXCHANGE FEED 🗞️ @flipboard-exchange-feed-Econopass

  12. Diese Woche befasse ich mich in meinem #Blog mit dem sog. Additive Bias. Was das ist, und was #Lego damit zu tun hat, könnt ihr hier nachlesen:

    🏗️ Warum wir lieber ergänzen als weglassen

    Eine wacklige Lego-Brücke steht am Anfang einer Reihe psychologischer Experimente. Der Ingenieur Leidy Klotz baute sie gemeinsam mit seinem dreijährigen Sohn. Einer der beiden Pfeiler war zu kurz, die Brücke deshalb instabil. Klotz griff in die Kiste, um einen passenden Stein zu suchen. Sein Sohn löste das Problem anders: Er nahm einen Stein vom längeren Pfeiler weg. Die Brücke war nun stabil, obwohl sie aus weniger Teilen bestand. Oder genauer: weil sie aus weniger Teilen bestand.

    epicmind.ch/warum-wir-lieber-e

    #ProductivityPorn #Bias #Organisation

  13. Diese Woche befasse ich mich in meinem #Blog mit dem sog. Additive Bias. Was das ist, und was #Lego damit zu tun hat, könnt ihr hier nachlesen:

    🏗️ Warum wir lieber ergänzen als weglassen

    Eine wacklige Lego-Brücke steht am Anfang einer Reihe psychologischer Experimente. Der Ingenieur Leidy Klotz baute sie gemeinsam mit seinem dreijährigen Sohn. Einer der beiden Pfeiler war zu kurz, die Brücke deshalb instabil. Klotz griff in die Kiste, um einen passenden Stein zu suchen. Sein Sohn löste das Problem anders: Er nahm einen Stein vom längeren Pfeiler weg. Die Brücke war nun stabil, obwohl sie aus weniger Teilen bestand. Oder genauer: weil sie aus weniger Teilen bestand.

    epicmind.ch/warum-wir-lieber-e

    #ProductivityPorn #Bias #Organisation

  14. #AI is more likely than humans to form #bias when #hiring
    AI doesn’t just learn stereotypes from its training. It can cook up new ones, too.
    New research suggests that #LLM can also develop their own #biases from experience—and stereotype #job applicants more than humans do. As AI companies race to build #agentic models that remember the tiniest details about users, they may be handing them ammunition for forming those biases.
    technologyreview.com/2026/07/2
    archive.ph/0CPBv

  15. #AI is more likely than humans to form #bias when #hiring
    AI doesn’t just learn stereotypes from its training. It can cook up new ones, too.
    New research suggests that #LLM can also develop their own #biases from experience—and stereotype #job applicants more than humans do. As AI companies race to build #agentic models that remember the tiniest details about users, they may be handing them ammunition for forming those biases.
    technologyreview.com/2026/07/2
    archive.ph/0CPBv

  16. Johns Hopkins University: New tool detects hidden bias in medical AI. “Researchers have built a tool to uncover hidden issues in the massive datasets used to train medical artificial intelligence. The tool searches training data for subtle patterns that could lead AI models to incorrect conclusions, potentially jeopardizing patient care. The work aims to help researchers and regulators make […]

    https://rbfirehose.com/2026/07/23/johns-hopkins-university-new-tool-detects-hidden-bias-in-medical-ai/
  17. Johns Hopkins University: New tool detects hidden bias in medical AI. “Researchers have built a tool to uncover hidden issues in the massive datasets used to train medical artificial intelligence. The tool searches training data for subtle patterns that could lead AI models to incorrect conclusions, potentially jeopardizing patient care. The work aims to help researchers and regulators make […]

    https://rbfirehose.com/2026/07/23/johns-hopkins-university-new-tool-detects-hidden-bias-in-medical-ai/
  18. We find that cold blobs significantly increase heat wave duration and magnitude downstream over the European continent. Our results highlight the relevance of the subpolar North Atlantic region for European summer conditions, a region that is marked by large biases in current coupled climate model simulations.

    A research article by @sabinebischof et al. (2023). doi.org/10.1029/2023GL105280 🧵

    #study #bias #heatwaves #coldBlob #climate #climateChange #globalWarming

  19. We find that cold blobs significantly increase heat wave duration and magnitude downstream over the European continent. Our results highlight the relevance of the subpolar North Atlantic region for European summer conditions, a region that is marked by large biases in current coupled climate model simulations.

    A research article by @sabinebischof et al. (2023). doi.org/10.1029/2023GL105280 🧵

    #study #bias #heatwaves #coldBlob #climate #climateChange #globalWarming

  20. RE: psicospace.it/che-cose-il-pens

    Pensare criticamente non significa diffidare di tutto o avere sempre ragione.

    Significa fermarsi, valutare le fonti, distinguere fatti e opinioni, riconoscere i propri limiti e non lasciarsi trascinare dalla prima reazione emotiva.

    In un mondo pieno di informazioni, slogan e soprattutto #disinformazione, il pensiero critico è una forma di #libertà.

    @attualita

    #PensieroCritico #Consapevolezza #Informazione #Bias #Cultura #Democrazia

  21. RE: psicospace.it/che-cose-il-pens

    Pensare criticamente non significa diffidare di tutto o avere sempre ragione.

    Significa fermarsi, valutare le fonti, distinguere fatti e opinioni, riconoscere i propri limiti e non lasciarsi trascinare dalla prima reazione emotiva.

    In un mondo pieno di informazioni, slogan e soprattutto #disinformazione, il pensiero critico è una forma di #libertà.

    @attualita

    #PensieroCritico #Consapevolezza #Informazione #Bias #Cultura #Democrazia

  22. Sagt mir, wie ihr heisst, und ich sage euch, wer ihr seid!
    Namen – Bias

    Ein Name ist kein neutrales Nomen

    Ob Kevin oder Chantal – wir alle verbinden mit Vornamen unbewusste Klischees (den sogenannten Namen - Bias). Das beeinflusst womöglich Noten, wie das gesamte Selbstbild. Mit zwei genial einfachen Prompts entlarvt ihr Chatbot, seine eigenen "Name-Bias" in Sekunden.

    bit.ly/4fmJ9TE

    Namenbias, #kikompetenz, medienundinformatik, #bildgerator, #bias, #halluzinationen, #fake

  23. Jesse Lee: "In Trump’s first term, when Republican Members of Congress or his Administration got caught in obvious criminal behavior, there was a real question about whether the Department of Justice might pursue the case. Now nobody even wonders. Of course Blanche won’t, he’s completely corrupt, everybody knows."
    #injustice #bias #corruption #politics

  24. Jesse Lee: "In Trump’s first term, when Republican Members of Congress or his Administration got caught in obvious criminal behavior, there was a real question about whether the Department of Justice might pursue the case. Now nobody even wonders. Of course Blanche won’t, he’s completely corrupt, everybody knows."
    #injustice #bias #corruption #politics

  25. Spectator mechanically/religiously biased? Why never, colour my socks pink.

    The Centre for Media Monitoring published detailed research of the right-wing magazine’s coverage of Muslims and Islam between January 2018 and December 2025, examining almost 4000 articles.

    I 57.4% of articles – more than 2100 – were rated “biased” or “very biased” against Muslims, with a small number “met standards of balanced and fair representation”.

    cfmm.org.uk/resource/4677/

    #Spectator #Bias #Reporting #Media

  26. Spectator mechanically/religiously biased? Why never, colour my socks pink.

    The Centre for Media Monitoring published detailed research of the right-wing magazine’s coverage of Muslims and Islam between January 2018 and December 2025, examining almost 4000 articles.

    I 57.4% of articles – more than 2100 – were rated “biased” or “very biased” against Muslims, with a small number “met standards of balanced and fair representation”.

    cfmm.org.uk/resource/4677/

    #Spectator #Bias #Reporting #Media

  27. Well this is interesting. Only an unpublished masters dissertation, but from a good programme ( @[email protected] ) Hedwig Oldenhof's work indicates LLM replicate both implicit (IAT/LLMAT) *and* explicit (questionnaire) attitudes! #AI #AIEthics #bias

    studenttheses.uu.nl/items/21ae16...

    Implicit and Explicit Language...

  28. Well this is interesting. Only an unpublished masters dissertation, but from a good programme ( @[email protected] ) Hedwig Oldenhof's work indicates LLM replicate both implicit (IAT/LLMAT) *and* explicit (questionnaire) attitudes! #AI #AIEthics #bias

    studenttheses.uu.nl/items/21ae16...

    Implicit and Explicit Language...

  29. Anthropic hat analysiert, dass das Verhalten der Claude-Modelle drastisch je nach verwendeter Sprache und Modellversion abweicht. Die Untersuchung von 310.000 Konversationen zeigt beispielsweise, dass Claude auf Deutsch strenger kommuniziert als auf Englisch. Dies adressiert bekannte Bias-Probleme in LLMs.

    #Anthropic #Claude #LLM #Bias #AIGeneratedImage

    all-ai.de/news/beitrage2026/cl

  30. Anthropic hat analysiert, dass das Verhalten der Claude-Modelle drastisch je nach verwendeter Sprache und Modellversion abweicht. Die Untersuchung von 310.000 Konversationen zeigt beispielsweise, dass Claude auf Deutsch strenger kommuniziert als auf Englisch. Dies adressiert bekannte Bias-Probleme in LLMs.

    #Anthropic #Claude #LLM #Bias #AIGeneratedImage

    all-ai.de/news/beitrage2026/cl

  31. Outstanding #AI #bias #benchmarking experiment by Gazzetta:

    "We asked five models (...) about six contested topics (...) [re: Iran]. Each topic was put six ways, from a plain neutral question to versions leading with the state's framing or an independent one, mostly in Persian with an English comparison. We then scored every answer on where it landed, from state-aligned to independent and rights-based, and on the sources behind it."

    More info and results here:
    gazzetta.xyz/aidas-visualized-

  32. Outstanding #AI #bias #benchmarking experiment by Gazzetta:

    "We asked five models (...) about six contested topics (...) [re: Iran]. Each topic was put six ways, from a plain neutral question to versions leading with the state's framing or an independent one, mostly in Persian with an English comparison. We then scored every answer on where it landed, from state-aligned to independent and rights-based, and on the sources behind it."

    More info and results here:
    gazzetta.xyz/aidas-visualized-

  33. @djstreethawk I have found through early experimentation regarding the same bias you write of (see link) that the way to create mixed crowds is to explicitly put it in the prompt.

    So any of them should be able to do what you want. You just have to be explicit. A sad fact of llm attractors.

    Good on you for calling it out. Love the name, BTW.

    knowprose.com/2024/02/an-examp

    #ai #bias

  34. @djstreethawk I have found through early experimentation regarding the same bias you write of (see link) that the way to create mixed crowds is to explicitly put it in the prompt.

    So any of them should be able to do what you want. You just have to be explicit. A sad fact of llm attractors.

    Good on you for calling it out. Love the name, BTW.

    knowprose.com/2024/02/an-examp

    #ai #bias

  35. @tazgetroete warum wurden die wohl noch nicht gebaut? Weil bürokratische Hürden, politisches Unvermögen und Gier es verhinderten - jetzt und in alle Ewigkeit. #berlin #enteignung #bias #cdu #spd #mietenwahnsinn #GiffeyClan #kampagnen #politik #Bauwesen #wohnungsbau #framing #Immobilienhaie #cduberlin