@olgakos
Cognition & Decision making. How humans interact with algorithms and AI, deal with uncertainty, and benefit from advice. Research applied to healthcare. Former SMDM President (smdm.org). https://profiles.imperial.ac.uk/o.kostopoulou
The School of Convergence Science at Imperial is inviting applications for fully funded PhD studentships on research projects that align with EPSRC’s remit.
For a project list and how to apply, see www.imperial.ac.uk/school-of-co...
Look out for project on Burnout in safety critical domains 😉
I saw it and reposted.
Please post this on LinkedIn
We are generally more forgiving of human than (an equivalent) algorithmic error. However, in contexts where using an algorithm is the norm, this effect is attenuated, dissipates or even reverses and we would rather retain the algorithm for future use.
www.cambridge.org/core/journal...
Super excited to share key findings from our study in JAMA NO!
Freq abx users (≥3 in 12mo) more likely to report:
-wanting abx for an URI 💊
-Needing help w/health-related materials 📖
-Medical maximizing (prefer to act vs watch/wait) 🚀
-Disbelief in science 🚫🔬
💻: jamanetwork.com/journals/jam...
What is explainable AI in medicine & what are the attributes of a good explanation? Based on a literature review and a Delphi study (I took part in this as member of the ExAIDSS Expert Group), this article provides an overview of the area and identifies gaps for future research. tinyurl.com/bddn8cs4
Very excited for today's invited speaker. @olgakos.bsky.social will be sharing her work on algorithms to support medical decision making.
hi Johanna, is this free to attend?
NB. "There is good evidence that partial knee replacement
offers better function, quicker recovery, and fewer
complications, as well as significant cost savings" compared to total knee replacement.
🆕 publication: In a behavioural experiment, where surgeons indicated their preference for total vs partial knee replacement, habitual practice (proportion of partial over all knee replacements in a year) was a stronger predictor of their preference than most clinical factors tested. How's your knee?
Scientists, given more time and flexibility, often stumble upon unexpected discoveries—like Alexander Fleming, who saw a mould had killed the bacteria he was growing, paving the way for antibiotics.
Recent study analysed such findings—more of them than you'd expect.
👉 www.nature.com/articles/d41...
Super interesting and the findings are not what you'd expect!
jamanetwork.com/journals/jam...
Or what it means to be a robot!
It’s science fiction of course but Ishiguro doesn’t see it like this (“Klara and the sun”)
This is a good blog about which AI to use now. www.oneusefulthing.org/p/which-ai-t...
I didn't even know such a thing (unstructured interviews) existed. Universities do not allow them but perhaps other industries do (?)
Cyclists! Please check the safety rating of your helmet on the Hiper website (www.hiperhelmets.org).
Study found "no link between the price of a helmet and the level of safety it provides". (Mine aren't doing very well...)
www.imperial.ac.uk/news/256314/...
If you still aim to submit a paper to our journal Health Literacy & Communication Open in December, please get in touch with me by the 20th at carissa.bonner@sydney.edu.au so we can plan how to manage your submission and open access fee waiver over the holiday break. Thank you!
An informative thread on AMR.
What was the gist? Just give me the gist! 😇
Predicting survival is fraught with uncertainty. This may not weigh much against giving patients choice but it is an interesting problem for those studying forecasting and preference elicitation. Does anyone know which 2023 study she is referring to (para 7)?
www.ft.com/content/241a... via @ft
I didn't know there was an interview. This doesn't surprise me. We have often found that the initial hypothesis determines the (process and) outcome of the diagnostic process. Also, see studies on predecisional information distortion.
That may simply be better use, better prompts. I'd be more interested in how the tool was used, how it impacted reasoning.
In any case, I think an interesting question is the implementation of LLMs in real-time clinical consultations (gradual information elicitation, belief updating).
Ah yes, someone else also sent me that. The structured format of responding (“reflective practice”) could have improved performance in the control group, so that the LLM group did not do significantly better.
An all-female panel and an inspiring discussion on climate change communications on the last day of the SJDM annual meeting, NYC. @sjdm-tweets.bsky.social
My editorial in the BMJ Q&S: measuring how individuals (here, clinicians) perceive and balance the benefits and costs of action can offer valuable insights into decision-making (here, in the context of low-value prescribing). qualitysafety.bmj.com/content/33/1...
Impressively cool attitude to arriving just before his talk and a clever way of 'buying time' by engaging with the audience, providing examples for his talk, while trying to put up his slides. Nice talk too. The whole session was great.
On changing the rules - not the players.
