“Gender- and sex-based discrimination in Google images: A comparative multimodal critical discourse analysis on the representation of same-sex couples”.

While it is undeniable that over reliance on AI-powered digital media has created a shift in epistemological understandings of how communication shapes, and is shaped, by the affordances of AI processes and power relations, it is equally true that the latest technologies function as multimodal resources for meaning making (Sindoni & Moschini, 2021) with the danger to offer a fertile ground for new forms of discrimination (Noble, 2018). The present study argues that AI-coded search engines, such as Google Images, function as semiotic resources where the interaction of the verbal and the visual modes communicates corporate interests and human bias that ultimately lead to structural gender- and sex-based discrimination. The aim of this study is to advance awareness on algorithmic discrimination by investigating asymmetric power relations and discriminatory frameworks in the combined visual/verbal representation of same-sex couples in Google Images from the vantage point of Multimodal Critical Discourse Analysis (Djonov & van Leeuwen, 2018). The comparative analysis concerns the Italian and the British digital landscape of Google Images, where same-sex couples are intended as a socio-culturally stigmatized group (Paterson & Turner, 2019). The examination uncovers different levels of structural gender- and sex-based discrimination in Google Images computations of verbal and visual resources, whose final output are thumbnail captioned images. The latter are intended as multimodal ensembles representing how same-sex couples epitomize gender- and sex-based discrimination in Google Images digital landscape.Since the results of the queries are highly context-dependent, the comparison of the Italian and the British digital landscapes encodes different discriminatory frameworks and power asymmetries that should prompt effective measures to contrast the perils of heteronormative bias in AI-based search engines.