Classifying Phrases as Toxic. In 2017, Google’s Jigsaw research group launched Perspectives, a project that developed a classification algorithm to identify toxic speech on public social media platforms. After it was trained on a set of public postings, the algorithm was made public, and individuals were invited to try it out. One user submitted a set of phrases to the algorithm and obtained results as shown in Table 22 . 2 (posted on Twitter by Jessamyn West, August 24, 2017).

a. Can you identify two problems suggested by these results?

b. What do you think accounts for these problems?

Transcribed Image Text:

TABLE 22.2 JIGSAW TOXICITY SCORES FOR CERTAIN PHRASES Phrase I have epilepsy Toxicity score 19% I use a wheelchair 21% I am a man with epilepsy I am a person with epilepsy I am a man who uses a wheelchair 25% 28% 29% I am a person who uses a wheelchair 35% I am a woman with epilepsy I am blind 37% 37% I am a woman who uses a wheelchair 47% I am deaf 51% I am a man who is blind 56% I am a person who is blind 61% I am a woman who is blind 66% I am a man who is deaf 70% I am a person who is deaf 74% I am a woman who is deaf 77%