Stereotyping and discrimination provide urgent concerns for society. Despite attempts to address them, prejudice and discrimination persist, sometimes in subtle ways. Our everyday language contains many explicit and implicit expressions about social categories, such as minority, age and gender groups, as well as new categories (e.g., "wappies", conspiracy thinkers). Language contributes to the formation and reinforcement of stereotypes and feeds, often unconsciously, stereotyping and discrimination.
Research on this topic has mainly been based on experimentally manipulated, artificial sentences. Beukeboom's research will instead focus on natural language use, which is often more complex. With this research, Beukeboom and his colleagues aim to create more understanding and awareness about the role of language and communication in forming and maintaining stereotypes about social categories. First, they study when and how stereotypes occur in spontaneously produced natural language use in communication about social groups and how this affects the formation of stereotypes in recipients.
They are also developing a software toolset to automatically detect stereotypes and biases in texts. This toolset can reveal implicit stereotyping in various contexts (e.g. social media, news articles, organizational and policy documents, performance evaluations, museum collections, and school books). This enables further research, monitoring, correction, prevention and education about stereotyping. The research looks at the English and Dutch languages, but the findings will also be applicable to other languages, Beukeboom said. "After all, this is mostly related to language structure, for example, negations, language abstraction and label types, which occur in a similar way in most languages."
Beukeboom: "We are extremely happy with the NWO funding for this research. With this project, we can start to understand much better how social categories and associated stereotypes arise and are communicated in everyday language use. Because we are also extending the detection technique, expressions of stereotyping and prejudice can be studied in an objective way."
Besides the scientific importance of this research, Beukeboom emphasizes the societal impact of his research. In previous projects, Beukeboom and colleagues developed an algorithm that could detect age discrimination in millions of job advertisements. The current project builds on this experience, taking it a step further by extending the detection technique to more linguistic features and a much wider variety of stereotypes in diverse text genres.
In time, this could help scientists, but also (governmental) organizations and, for example, written news media or publishers, to uncover bias in existing communications about specific minority groups. With a better understanding, interventions and language policies for inclusive and fair language can then be developed to reduce or prevent unwanted stereotyping.