Monday 24 April 2017


Controversially, German immigration authorities have announced plans to utilise speaker- and speech-recognition software to screen refugees applying for asylum. Automated analytics can help expedite the applications of individuals and families from war-zones but could also as expeditiously disqualify those that the algorithm determines don’t have the speech markers and regionalism that corroborate claims that they are fleeing from Syria but are rather seeking better economic opportunities or welfare.
Having used voice samples and linguistic analysis since the 1990s to screen refugees, the idea is not a new one and human experts have demonstrated a high error rate and many worry that those misjudgements might only be magnified by automation. Affecting a semester-abroad accent is one thing but native speakers are also apt to adopt the expressions and pronunciation of those around them unconsciously out of deference fairly swiftly. Though immigrants have people advocating for them and arguing that the process is prone to error, but I think Germany’s plan has elicited less apprehension than a similar process might soon judge our fitness for employment. Both potential applications are of great consequence and deserve equal scrutiny.