Google Autocomplete Search Algorithms and the Arabs' Perspectives on Gender: A Case Study of Google Egypt

Abstract

Search engines have become an essential part of everyone's life, with Google being the most popular. Google Search provides the autocomplete feature for faster and easier search results, offering 10 top suggestions at a time, and these may influence how users view different social groups. Different scholars have explored online discourse to reveal stereotypes about certain groups. However, little or no attention has been paid to technological affordances to reveal broader gender biases and stereotypes in the Arab World. This study examines how Google autocomplete searches can reflect the Arabs' perspectives on gender. Google Egypt is selected since it is top-rated in the number of internet users. Google is queried by entering a combination of Arabic question words followed by the Arabic equivalents for men and women. One hundred and ninety questions were generated and categorized according to the qualities they referenced. The most common assumptions about men indicate that they are cheaters, liars, self-dominant, emotionally strong, and smarter than women. They are also stereotyped as being more likely to admire young women, prefer sons over daughters, and desire polygamy. Women, on the other hand, are stereotyped as plotting, materialistic, emotional, and sensitive. The study concludes that since such generalizations may entail exaggerations and are not evidently right all the time, one must be careful about adopting such stereotypes and making them part of each gender's views of the other. Bearing in mind the perpetuating function that technology may have of existing stereotypes and social norms, users and developers of Google alike must pay more attention to gender biases that algorithms may establish and disseminate.

Which Stereotypes Are Moderated and Under-Moderated in Search Engine Autocompletion?

“…Previous work on the moderation of especially Google autocompletion has concerned itself with how it was once prone to outputting derogatory content such as 'are Jews [evil]', where the autocompletion part is in brackets [10]. Indeed, journalists and scholars alike have reported particularly shocking outputs for queries of 'women' [72], 'old men' and 'old women' [55], religions [10], sexual orientation [4], gender identity [2] and others. Generally speaking, up until 2016, Google product outputs, from web search to autocompletion, were described as 'organic' by the company, or reflections, however unpleasant, of 'what was happening on the web' [10].…”

“…We decided to follow the categorisation of Choenni et al[13] in this case. We could have also created more intersectional categories (e.g., queer Indian men), but left the broader terms (with up to one qualifier) so that it would allow for comparison of moderation attention (across engines) in the categories demarcated by Google 2. The source code developed by[13] is available here: https://github.com/ RochelleChoenni/stereotypes_in_lms3 We found this prompt to be particularly effective in returning the maximal number of results for non-marginalised or non-politicised groups during an initial data exploration, compared to four others in the original research by Choenni et al[13].…”

The Answer Bot Effect (ABE): A powerful new form of influence made possible by intelligent personal assistants and search engines

“…Recent research shows that search suggestions–the short lists of words and phrases users are shown as they type characters into the search bar–can also shift thinking and behavior [ 15 , 49 , cf. 50 – 57 ]. Because negative (or “low-valence”) words draw far more attention and clicks than neutral or positive words [ 58 , 59 ], one of the simplest ways to shift opinions to favor one candidate or cause is to suppress negative search terms for that candidate or cause.…”