Probabilistic Images: the composite as a space of possibility

Maria Dada and Nishat Awan presented at the Figurations: Persons In/Out of Data conference.

January 11th at 7:34pm

A geographic information system (hereafter GIS) is a computer system for capturing, storing, checking, and displaying data related to positions on the earth’s surface. It makes use of remote sensing and photogrammetry in order to recreate the earth’s surface as an image. The so-called ‘absolute accuracy’ of the remote sensed image depends on the resolution of the geolocated source data. However, rather than, as the term suggest, produce absolute truths or falsehoods, ‘absolute accuracy’ is a probabilistic relationship to the ground truth that the GIS algorithms themselves produce. It is in that sense not at all absolute but rather probabilistic.

As such, the designers of the algorithms in conversation with the visual regimes of the border ‘tune’ them to produce the desired outcome without necessarily being aware of the ground upon which the algorithm is making its calculations. Their process is one of trial and error in relation to what political geographer Louise Amoore refers to as an aperture whose size and shape the programmer unknowingly adjusts. This paper seeks to find ways that would somehow reveal the instability of ‘absolute accuracy’ and reintroduce doubt; as Amoore states within ‘contemporary machine learning algorithms, doubt becomes transformed into a malleable arrangement of weighted probabilities.’

Image credit: