Machines, Parents, and Differences

David Barner’s “Accessing the unsaid: The role of scalar alternatives in children’s pragmatic inference” and Alex Stiller’s “Ad-hoc scalar implicature in adults and children” both speak on implicature and its presence among children. Barner explores young children’s inability to compute scalar implicatures— defined well by Stiller: “scalar implicature refers to the conversational shorthand of using weak terms to imply the negation of stronger ones that lie along the same scale” (1). Barner focuses most on the scale of <a, some, many, all, etc>. Children were found to lack the knowledge of relevant scalar alternatives. Stiller focused more on ad-hoc implicatures which means that the “scale arises (or fails to arise) from the real-world context rather than the lexical items” (1). In 3 experiments, he found that children were able to use ad-hoc scales, which seems to suggest that young children have the underlying functions needed to compute implicatures. 

One thing I found very interesting, was Barner’s discussion of the fact that intended meanings go beyond the literal meanings of words. When I look back to class activities and readings we did related to audio recognition like Siri, this becomes very relevant. If we want machines in the future to become as intelligent as possible, they need to be able to compute scalar implicatures. Children incorrectly interpret commonly used language and do not receive the full intent of the word “some” compared to “all”. Computers, likewise, should have this ability. However, as Barner explains, this extends to many other things. Computers, if we wish them to truly be an assistant and become human-like, need to understand intended meaning and not just a dictionary definitions. A lot of research should go into how we give machines this ability to better understand humans. 

This reading also reminded me of Carnie’s work we discussed very early in the quarter. I remember Carnie explaining that parents’ correcting their children actually showed no tangible impacts on their learning— which I still find hard to believe. This reading brings that to light again and makes me wonder at what point do we finally acquire the complete ability to compute scalar implicatures. 

Even further, do non native speakers have no problem computing scalar implicatures across languages, or are the process and scales different across languages? For example, if one language had more diversity and words along a scale, does computing scalar implicatures become more difficult in a new language by the lack of descriptive words? I think there is much to be discussed in how this seemingly small subset of language is emblematic of larger themes we have covered in this course.