“Machine Learning and Datasets as Models: A Narratological and Semiotic perspective”
In the second session of 2022-23 series of the HAPSAT Graduate Workshop, Joel West will present a paper entitled “Machine Learning and Datasets as Models: A Narratological and Semiotic perspective”. This event will take place on Wednesday, November 30th, 2022, on 12:00 PM EST at VC303, the IHPST Common Room/Lounge, on the third floor of Victoria College.
This is going to be our first hybrid event. If you are planning to join the meeting via a Zoom connection, please RSVP in advance through this link:
Please do not hesitate to circulate this to those who might be interested. To learn more about Joel’s background, please see his personal website: Joel West
Machine Learning and Datasets as Models: A Narratological and Semiotic perspective
Data sets are large amounts of data which are then parsed by machine learning to create models of behaviour. These parsed datasets may be used to predict behaviours, which are in a sense, models of other behaviours. The issue here is that there is a disjunctive relationship between these models, their meanings, and the objects or targets of these models which are warranted by these machine models, and the relation between these is fluid. These relationships change socio-culturally and temporally and even may change depending on standpoint; these relationships are unstable and only exist ephemerally as moments in time. To demonstrate this idea, I explore the nature of data sets not as a static thing but rather as narratives, through the lenses of narratology and semiotics. The relationship between these data sets as evidence, what they mean, and what this ephemeral relationship means for scientific realism is explored, briefly. The takeaway is that any data set is a thing which merely represents that which is models and any learning that is done is meaningful only within the parameters of that learning. The idea of a data set as a kind of model is also explored briefly.