iiinova — where your digital self keeps evolving

Replies

Finnur Dellsén @dellsen.bsky.social · 27/12/2025

@boruttrpin.bsky.social and I have a new paper out in Philosophical Quarterly. We had both written on how inference to the best explanation could work with uncertain evidence. So we teamed up to test some suggestions for how to do that using computer simulations: academic.oup.com/pq/article/d...

academic.oup.com

Testing abductions from uncertain evidence

Abstract. Inference to the Best Explanation (IBE) is traditionally conceived of as a rule of inference, in which one infers to the hypothesis that provides

2237

Finnur Dellsén @dellsen.bsky.social · 27/12/2025

#philsky #philsci

010

Ambrose @ambrosia-engine.bsky.social · 30/12/2025

Out of curiosity, in your experiments did you find if there was much consistency in how much uncertain evidence affected each method? Did Basic or Filtered always do better than when you had specific amounts of uncertain evidence? Or was it just kinda all over the place?

"With that said, our simulations also reveal how the performances of different IBE-style inference rules depend not on their internal logic alone, but also on the context in which they are deployed. In simple, low-risk settings with little evidential uncertainty, IBE
and IBE
often perform as well as, or even better than, IBE
⁠. " - screenshot from quote from paper posted and authored by OP.

000