OneMonkey raised the issue of voice recognition changing people’s accents and vocabulary, as they modify their speech in order to be recognised by their gadgets. As I don’t have a smart phone and can’t imagine giving voice commands to inanimate objects I don’t feel particularly qualified to comment. I only have a couple of bits of anecdotal evidence about people with strong accents not being understood by phones and tablets, and a particularly trying afternoon where I failed to make the HMRC automated phone service understand a single phrase I said (I suspect they do it on purpose – it is entirely possible that the big companies we see as tax-dodgers have tried to register properly but been thwarted by the byzantine complications of HMRC bureaucracy). None of the situations I’m thinking of resulted in anyone modifying their speech, they just gave up on voice recognition.
Presumably voice recognition on, say, phones is calibrated to a particular mode of speech and pre-programmed with a certain vocabulary. If you stray too far outside either of those it will struggle. Anyone who tries it and fails spectacularly is likely to give up on it as unsuitable, but I imagine if it mostly recognises what you say except for a few words you might try modifying those words until it could handle them. I modify written words, both vocabulary and sentence structure, for texting and tweeting. Naturally, I’m quite verbose (no need to be polite, I know you’ve noticed) and I often write a tweet the way I want to say it then rearrange, shorten, replace ‘placed’ with ‘put’ etc to get within my character limits. I’m not aware of it spilling over into other areas of life, however, and with written words I think it’s easier to keep separate vocabulary pots. In speech it’s easier to get into habits, and if you start pronouncing a word slightly differently you may find that becomes the way you unconsciously do it. The two areas I can think of it happening are where a) a particular accent pronounces the same word very differently from the standard (e.g. bus) or b) an accent doesn’t clearly distinguish between words that the standard does differentiate (e.g. look, luck). It would be fascinating to see how the widespread use of voice recognition changes natural speech over the next decade.