Based on my understanding about how Merlin Sound ID works, in my opinion the answer to your question is "no". You are wise to not blindly trust that Merlin Sound ID found a real Cackling Goose in the flock of geese that you had near your house. There are two aspects of how Merlin Sound ID works that are behind my opinion:
1. Merlin Sound ID scans through recordings using 3-second-long subsets of an entire recording, and within each of these 3-second clips, it assigns each vocalization that it can identify to its most likely species. So, all it would take is for one Canada Goose to give a slightly "off" vocalization a single time, or just the right interference from background noise (the microphones are phones pick up all sorts of background noise) and "Cackling Goose" could pop up as being identified. You didn't say whether you keep seeing Cackling Goose being repeatedly identified, or just a single time during the time that Merlin Sound ID was running. I would tend toward skepticism of any ID that was made a single time during the period that Merlin Sound ID is running. 2. Behind the curtain, Merlin Sound ID isn't assigning a single species to each vocalization that it picks up. Instead, for every separate vocalization, Merlin Sound ID is calculating how close that vocalization is to resembling every species for which the Sound ID model has been trained. So, it is possible that Sound ID will consider two species to be very similar in likelihood, but the Merlin app only shows you the most likely species regardless of whether the most likely, and the second most likely, species might both have a very high likelihood of being the species that made the vocalization. In other words, the presentation of information from the Merlin Sound ID model can lead to a false impression of certainty in an identification. Basically, as with interpretation of output from any machine-learning model, human intelligence is required in order to assess whether or not artificial stupidity has occurred. Wesley Hochachka ________________________________ From: bounce-128062900-3494...@list.cornell.edu <bounce-128062900-3494...@list.cornell.edu> on behalf of Marty Schlabach <m...@cornell.edu> Sent: Saturday, March 2, 2024 12:29 To: CAYUGABIRDS-L <cayugabird...@list.cornell.edu> Subject: [cayugabirds-l] geese One evening earlier this week, a fairly large flock of geese were grazing in field well behind our house. It was almost dark. I couldn’t see them, but could ID Canada and Snow Geese. I opened Merlin and it IDed Canada, Snow and Cackling Geese. Can I believe Merlin’s ID of the Cackling Geese? --Marty =========================================== Marty Schlabach m...@cornell.edu 8407 Powell Rd. home 607-532-3467 Interlaken, NY 14847 cell 315-521-4315 =========================================== -- Cayugabirds-L List Info: Welcome and Basics<http://www.northeastbirding.com/CayugabirdsWELCOME.htm> Rules and Information<http://www.northeastbirding.com/CayugabirdsRULES.htm> Subscribe, Configuration and Leave<http://www.northeastbirding.com/CayugabirdsSubscribeConfigurationLeave.htm> Archives: The Mail Archive<http://www.mail-archive.com/cayugabirds-l@cornell.edu/maillist.html> Surfbirds<http://www.surfbirds.com/birdingmail/Group/Cayugabirds> ABA<https://www.aba.org/birding-news/> Please submit your observations to eBird<http://ebird.org/content/ebird/>! -- -- (copy & paste any URL below, then modify any text "_DOT_" to a period ".") Cayugabirds-L List Info: NortheastBirding_DOT_com/CayugabirdsWELCOME_DOT_htm NortheastBirding_DOT_com/CayugabirdsRULES_DOT_htm NortheastBirding_DOT_com/CayugabirdsSubscribeConfigurationLeave_DOT_htm ARCHIVES: 1) mail-archive_DOT_com/cayugabirds-l@cornell_DOT_edu/maillist_DOT_html 2) surfbirds_DOT_com/birdingmail/Group/Cayugabirds 3) aba_DOT_org/birding-news/ Please submit your observations to eBird: ebird_DOT_org/content/ebird/ --
