Visually Grounded Speech Models have a Mutual Exclusivity Bias
This work addresses the problem of understanding cognitive biases in AI models for researchers in computational linguistics and cognitive science, though it is incremental as it extends existing bias studies to continuous speech inputs.
The study investigated whether visually grounded speech models exhibit a mutual exclusivity bias, similar to children, when learning from natural images and continuous speech audio, finding that the bias is present and stronger in models with more prior visual knowledge.
When children learn new words, they employ constraints such as the mutual exclusivity (ME) bias: a novel word is mapped to a novel object rather than a familiar one. This bias has been studied computationally, but only in models that use discrete word representations as input, ignoring the high variability of spoken words. We investigate the ME bias in the context of visually grounded speech models that learn from natural images and continuous speech audio. Concretely, we train a model on familiar words and test its ME bias by asking it to select between a novel and a familiar object when queried with a novel word. To simulate prior acoustic and visual knowledge, we experiment with several initialisation strategies using pretrained speech and vision networks. Our findings reveal the ME bias across the different initialisation approaches, with a stronger bias in models with more prior (in particular, visual) knowledge. Additional tests confirm the robustness of our results, even when different loss functions are considered.