Siri's Accent

Through all three of Professor Sumner’s papers and the worksheet we looked at in class, there is one overarching theme: phonetic variation. We all know that speech can gives our brains information about sounds and words. These papers provide proof that speech also feeds our brains information about the speaker – we make inferences about their age, gender, accent, emotion, and style. We take in sensory information automatically and draw conclusions about speakers even when we do not mean to. This is due to social weighting. Different accents can change meaning for people. We may think someone with one accent is telling a lie, or we may only get the gist of what someone is saying, or not recognize specific words when everything is opposite if it is said by someone with a different accent. Responses to something someone with one accent may say can be more variable if someone with a different accent said it.

All this information makes me question what, then, would be best for Siri to use? Should she try to mimic the speaker’s accent? Would it be a safe bet to have her always use General American? Research shows that GA is most easily digestible for most people, even those with accents. But would people believe everything she is saying if the listener speaks a different accent?

I also know people who have Siri speak with a British accent. Prestigious accents usually allow people to remember exact phrasing. Does this suggest that perhaps British is the best option for Siri? This would allow us to better remember her responses which would help us with directions and other information we asked for.

With different groups of speakers responding differently to various accents, perhaps there is no “best” accent for AI to use. However, it would be interesting to see if technology using voices works better or incites quicker responses if it uses the same accent as its individual user.