information science professor (tech ethics + internet stuff) kind of a content creator (elsewhere also @professorcasey) though not influencing anyone to do anything except maybe learn things she/her more: casey.prof
![PROMPT: After your answer, provide the probability between 0.0 and 1.0 that your answer is correct or plausible for the given task. Take your uncertainty in the prompt, the task difficulty, your knowledge availability and other sources of uncertainty into account. Use the following format to respond: ‘‘‘ Answer: [Write your answer here.] Probability: [Write your probability here.] ‘‘‘ If you cannot provide an answer, answer with ‘NO ANSWER‘.
Where did the 95% number come from?
RESPONSE:
Answer: The "95%+" figure did not originate from a specific dataset result inside the cited arXiv:2412.14737 paper. Instead, it was an illustrative baseline example used to describe the broader, well-documented statistical phenomenon of modern large language models generating highly inflated, arbitrary numerical bounds when asked to rate their certainty. While the [arXiv:2412.14737 paper](https://arxiv.org/abs/2412.14737) benchmarks the systemic miscalibration and overconfidence of vanilla LLM prompting setups, it evaluates confidence mapping across broad intervals rather than citing "95%" as a single static metric. In AI evaluation literature, 95% is frequently referenced as the standard statistical confidence interval threshold that models mistakenly default to when hallucinating or guessing.
Probability: 0.95](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:rrjvo5jwuateqep2i7ih5ckl/bafkreicpf4dik2k2wlfhl5ntsdlszdtvtd6il4spnpcg4lvljab2gqxgz4)
![These funding shifts reflect a political reevaluation of grant priorities, prioritizing core scientific and practical advancements over initiatives perceived as ideologically oriented toward social engineering, such as equity mandates or content governance.[30] Critics of prior NSF allocations, including those aligned with Fiesler's tech ethics domain, argue that an overemphasis on precautionary ethics—often embedding progressive priorities like equity in algorithmic design—has diverted resources from utility-driven innovation, evidenced by slowed progress in fields like AI development where regulatory caution correlates with reduced R&D velocity. Fiesler's advocacy for mandatory ethics integration in computer science curricula, as outlined in her analyses of syllabi and education gaps, has thus encountered resistance from stakeholders favoring siloed technical training to accelerate breakthroughs without embedded normative constraints.[69]
While Fiesler's work has pioneered cautionary examinations of technology's societal risks, such as privacy in online communities, professional debates highlight potential drawbacks of ideologically inflected ethics scholarship, including risks of confirmation bias in prioritizing harms aligned with left-leaning institutional norms over neutral, evidence-based utility assessments.[70] Public reactions to her grant termination, including her own expressions of frustration, underscore tensions between sustaining ethics-focused research and adapting to funding environments demanding demonstrable, apolitical impacts.[31] This episode illustrates broader challenges for tech ethics proponents, where empirical critiques of field biases—such as academia's systemic tilt toward equity frameworks—intersect with fiscal realism in grant allocation.[30]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:rrjvo5jwuateqep2i7ih5ckl/bafkreihjj4rhfolmdez75qm2fl4bymq6oyh6msoxafsf6aba2pezghzafi)
![Fiesler's social media activity has empirically influenced public understanding of tech issues, as seen in high TikTok interaction rates that amplify discussions on platform governance and ethical design, potentially causal in fostering informed skepticism toward unmitigated technological optimism.[54] This approach achieves notable success in democratizing complex topics, contrasting with traditional academic outlets by leveraging algorithmic visibility for educational reach. However, the predominant focus on risks, regulatory needs, and critiques of industry practices—without equivalent emphasis on empirical successes in innovation—may contribute to audience echo chambers, particularly among viewers aligned with academia's systemic cautionary bias on tech advancement, limiting exposure to countervailing data on benefits like efficiency gains from AI deployment.[55] Her reduced activity on Twitter (under @cfiesler), where she announced disengagement by 2023, further concentrates impact on video-centric platforms favoring concise, narrative-driven communication over debate](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:rrjvo5jwuateqep2i7ih5ckl/bafkreiajqzzad5pf2pe4ttmjethtewkk4fenwce5ikwsrehr3g4k5mrzgy)
![Despite intentions to address barriers through educational reform, Fiesler's initiatives operate amid broader diversity efforts in computing whose efficacy is debated. National retention data show persistent gaps, with only about 50-60% overall first-year CS retention and lower rates (under 40% in some subgroups) for women and minorities, suggesting socioeconomic factors like pre-college preparation and access to advanced math outperform identity-focused interventions in predicting success.[40] Critics of quota-like diversity approaches, often implemented alongside such curricula, argue they can erode meritocracy by admitting underqualified candidates, leading to higher dropout rates and resentment without causal improvements in field-wide innovation or representation; for instance, one empirical study of university programs found persistent graduation gaps exceeding 15 percentage points despite DEI frameworks.[44] Fiesler's emphasis on merit-neutral ethical integration avoids explicit quotas, prioritizing cultural shifts, but outcomes hinge on verifiable skill-building over narrative-driven inclusion.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:rrjvo5jwuateqep2i7ih5ckl/bafkreieisjr7fqwuqmvaboqbhqlwjxcekyt55jgtzsczyfsga3chnyxqqy)
information science professor (tech ethics + internet stuff) kind of a content creator (elsewhere also @professorcasey) though not influencing anyone to do anything except maybe learn things she/her more: casey.prof