Assessing gender bias in machine translation: a case study with Google Translate

By Marcelo O. R. Prates, Pedro H. C. Avelar, L. C. Lamb. 2018

403 citations recorded by the source corpus.

Recently there has been a growing concern in academia, industrial research laboratories and the mainstream commercial media about the phenomenon dubbed as machine bias, where trained statistical models—unbeknownst to their creators—grow to reflect controversial societal asymmetries, such as gender or racial bias. A significant number of Artificial Intelligence tools have recently been suggested to be harmfully biased toward some minority, with reports of racist criminal behavior predictors, Apple’s Iphone X failing to differentiate between two distinct Asian people and the now infamous case of Google photos’ mistakenly classifying black people as gorillas. Although a systematic study of such biases can be difficult, we believe that automated translation tools can be exploited through gender neutral languages to yield a window into the phenomenon of gender bias in AI. In this paper, we start with a comprehensive list of job positions from the U.S. Bureau of Labor Statistics (BLS) and used it in order to build sentences in constructions like “He/She is an Engineer” (where “Engineer” is replaced by the job position of interest) in 12 different gender neutral languages such as Hungarian, Chinese, Yoruba, and several others. We translate these sentences into English using the Google Translate API, and collect statistics about the frequency of female, male and gender neutral pronouns in the translated output. We then show that Google Translate exhibits a strong tendency toward male defaults, in particular for fields typically associated to unbalanced gender distribution or stereotypes such as STEM (Science, Technology, Engineering and Mathematics) jobs. We ran these statistics against BLS’ data for the frequency of female participation in each job position, in which we show that Google Translate fails to reproduce a real-world distribution of female workers. In summary, we provide experimental evidence that even if one does not expect in principle a 50:50 pronominal gender distribution, Google Translate yields male defaults much more frequently than what would be expected from demographic data alone. We believe that our study can shed further light on the phenomenon of machine bias and are hopeful that it will ignite a debate about the need to augment current statistical translation tools with debiasing techniques—which can already be found in the scientific literature.

Subfields: Artificial Intelligence, Language and Linguistics, Natural Language Processing, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Language and Linguistics

In Natural Language Processing

In Safety, Robustness & Interpretability

Citations in the corpus

Language (Technology) is Power: A Critical Survey of “Bias” in NLP (2020, neutral): “…Dinan et al. (2019); Florez (2019); Jumelet et al. (2019); Lauscher et al. (2019); Liang et al. (2019); Maudslay et al. (2019); May et al. (2019); Prates et al. (2019); Prost et al. (2019); Qian et al. (2019); Swinger et al. (2019); Zhao et al. (2019); Zhou et al. (2019); Ethayarajh (2020);…”
The Woman Worked as a Babysitter: On Biases in Language Generation (2019, background): “Recent works in machine translation (Prates et al., 2018) and dialogue systems (Henderson et al., 2018) have brought to attention the perpetuation of biases in natural language generation (NLG) systems.”
Societal Biases in Language Generation: Progress and Challenges (2021, background): “Furthermore, Prates et al. (2019) and Farkas and N´emeth (2020) compare pronoun gender biases in translations (induced with prompts) to real-world statistics.”
Beyond English-Centric Multilingual Machine Translation (2020, methodology): “Further, work has analyzed the presence of bias in automatic translation systems (Prates et al., 2019) as well as the use of translation systems in various applications such as medicine (Vieira et al.”
Toward Gender-Inclusive Coreference Resolution (2019, background): “, 2019; Gonen and Goldberg, 2019), sentiment analysis (Kiritchenko and Mohammad, 2018), machine translation (Font and Costa-jussà, 2019; Prates et al., 2019; Dryer, 2013; Frank et al., 2004; Wandruszka, 1969; Nissen, 2002; Doleschal and Schmid, 2001), among many”
Hi, my name is Martha: Using names to measure and mitigate bias in generative dialogue models (2021, background): “, 2021), neural machine translation (Cho et al., 2019; Stanovsky et al., 2019; Prates et al., 2020; Renduchintala et al., 2021; Savoldi et al., 2021), or language modeling (Nangia et al.”
Does Gender Matter? Towards Fairness in Dialogue Systems (2019, background): “In the work (Prates et al., 2018), it is shown that Google’s translation system can suffer from gender bias by making sentences taken from the U.”
Mitigating Gender Bias in Natural Language Processing: Literature Review (2019, background): “Current MT models predict the source to be male a disproportionate amount of time (Prates et al., 2018; Vanmassenhove et al., 2018).”
Equalizing Gender Bias in Neural Machine Translation with Word Embeddings Techniques (2019, neutral): “Using pre-trained embeddings for initialization seems to improve the translation is slightly coherently with previous studies (Qi et al., 2018).”
Towards Controllable Biases in Language Generation (2020, background): “evious work deﬁne bias metrics (Sheng et al., 2019; Bordia and Bowman,2019),createdatasets forthe study of biases (Sheng et al., 2019; Nadeem et al., 2020), and analyze biases in machine translation (Prates et al., 2019; Stanovsky et al., 2019). The major challenge of analyzing and mitigating biases”
A Distributional Approach to Controlled Text Generation (2020, background): “There has been a large body of work analysing these distributional biases (Blodgett et al., 2020; Stanovsky et al., 2019; Prates et al., 2020; Sheng et al., 2019a; Brown et al., 2020b).”
Predictive Biases in Natural Language Processing Models: A Conceptual Framework and Overview (2019, background): “However, Preotiuc-Pietro et al. (2015) showed that these classifiers primarily picked up on differences in age and/or gender – those with PTSD were more likely to be older.”
Measuring and Reducing Gendered Correlations in Pre-trained Models (2020, neutral): “…et al., 2019; Romanov et al., 2019), sentiment analysis (Kiritchenko and Mohammad, 2018), coreference resolution (Rudinger et al., 2018; Zhao et al., 2018a; Webster et al., 2018), image captioning (Zhao et al., 2017), and machine translation (Stanovsky et al., 2019; Prates et al., 2018).”
On Measuring Gender Bias in Translation of Gender-neutral Pronouns (2019, methodology): “Most recently, Prates et al. (2018) utilized a heat map regarding occupation and adjectives in 12 languages to evaluate Google translator.”
Automatic Gender Identification and Reinflection in Arabic (2019, methodology): “Prates et al. (2018) used Google Translate to translate a set consisting of a list of jobs and gender-speciﬁc sentences from a variety of gender-neutral languages into English.”
Reducing Gender Bias in Neural Machine Translation as a Domain Adaptation Problem (2020, positive): “Each profession is from the list collected by Prates et al. (2019) from US labour statistics.”
Investigating Failures of Automatic Translation in the Case of Unambiguous Gender (2021, methodology): “Prates et al., 2019) are reflected by the percentage of examples for which the MT system chooses the gender-stereotypical pronoun as opposed to the anti-gender-stereotypical.”