Attempts are made to mitigate bias by introducing artificial intelligence (AI). Technol- ogy however, often acts as a double-edged sword, having the abilities to both identify and mitigate bias, or amplify inequality, reinforce existing stereotypes and increase injustice.

Recognising both the potential but also the limitations of AI, this thesis explores the idea of reflexive designer-AI interactions, as a new form of human-machine interac- tion towards more reflective design practitioners who are able to surface, dismantle and re-think personal and collective imaginings.

Following a speculative and introspective research through design approach, this thesis explores such reflexive interactions in the context of gender representation in child toys. The hypothesis is that the introduction of reflection and a change in mind- set when engaging with AI, can be productive in terms of mitigating gender bias in child toys.

Technological exploration insights are translated into four design tactics, which are applied and explored in practice, by designing three gender-ambiguous child toys. Each toy represents a new reflexive design-AI workflow. Each workflow differently illustrates human and non-human collaboration that surfaces, defamiliarizes and dis- mantles personal and collective imaginings of gender in toys.

Taking into account the insights from the experiments, as well as prototype testing with children and evaluating expert interviews, this project concludes that reflexive interactions are potentially productive to surface, dismantle and re-familiarize per- sonal bias and collective imaginings. Furthermore, does this thesis suggest that AI’s often negatively described behaviour like confusion and inconsistency, also carry the power to trigger reflective practices that help surfacing and challenge bias.