this piece on covid voice detection and other forms of vocal diagnostics is really worrying. https://www.nature.com/articles/d41586-020-02732-4#ref-CR6

(1) it takes highly dubious claims about these tools' accuracy at face value eg that AI can detect depression from voice samples with '94% accuracy' (but there are similar claims re eg PTSD, Alzheimers, autism and lots of other neurodevelopmental and neurodegenerative disorders)

(2) at the very least, this kind of thing needs A LOT more context. psychologists can't even detect depression with 94% accuracy. what is meant by depression? since the definition in the DSM is variegated, regularly updated, and differently diagnosed in different places/cultures

(3) the lack of context/critique in relation to 'success' stories' like depression detection matters because it makes it seem like covid voice diagnosis is just around the corner, when there's no evidence of that at all. just a few scientists posing this future as inevitable

(4) and this matters because companies like Sonde Health (who are discussed in the article) are already marketing products claiming to be able to DO covid voice diagnosis for employers to integrate into 'safe return to work' systems (which is not raised) https://www.sondehealth.com/sondeone-page 

(5) as a result, the article effectively reads as technosolutionist advertorial for these kinds of companies and the scientists who are associated with or, in some instances, own them

(6) no surprise then that the piece's account of 'risk' runs for 3 short paragraphs at the end, and is limited to privacy, consent, insurance discrimination, and 'false positives', none of which are situated in relation to broader political systems.

(7) there is no account of big tech, capital, control, surveillance, security, policing, the defunding of public health, or any other oppressive structures such tools would emerge into. no discussion either of whether/how this stuff shades into physiognomy (it does).

(8) so that finally the emergence and automation of voice diagnostics is presented as not only inevitable but 'a good thing' where the only question is "what’s going to happen when we put it into practice.” not if. when.

(9) a reminder that this is in Nature. supposedly a/the flagship science journal.

(10) for a much better, more interesting and politically aware take on covid voice diagnostics, see @seandockray's recent performance for Machine Listening @unsound @liq_arch @sternjoelstern

a few more resources on covid voice diagnostics and other related themes in machine listening available here https://machinelistening.exposed/topic/listening-with-the-pandemic/