Using machine vision to explore Instagram’s everyday promotional cultures
Social media platforms like Instagram are at the centre of the largest commercialisation of public life and intimate experience in history. They dramatically expand the access marketfers have to our cultural practices and experiences.
We have been exploring the interplay between our creative use of Instagram to turn our life into flows of images and the translation of those images into data.
Platforms use our images to train ‘machine vision’ systems that classify faces, expressions, patterns, objects, and brand logos. These classifications are used to train automated recommendation systems that tailor content and advertising.
From 2020 we are working on an Australian Research Council Discovery Project called ‘Using machine vision to explore Instagram’s everyday promotional cultures’.
In this project we’re going to explore Instagram’s participatory and promotional cultures and experiment with how we might simulate the capacity of algorithms to classify and cluster the flows of images we create and circulate.
The project team includes Associate Professor Daniel Angus, Professor Jean Burgess and Associate Professor Nicholas Carah.
The video below is a talk that presents some of our beginning arguments and explorations of Instagram using machine vision.
We want to understand the nuanced interplay between our creative expressive practices and the data-processing power of platforms. To do that we’re going to combine well-honed approaches to the participatory cultures of digital media platforms — talking to users and analysing the content they create — with novel techniques for exploring how machine vision systems work.
We will explore everyday practices on Instagram, talk to producers and users, and track the strategies of marketers and platforms. Alongside this, we identify the ‘state of the art’ machine vision approaches, scrape datasets of Instagram images, and then experiment with what these algorithms can ‘do’ with the images.
We’ll use this site to document our project. As a starting point some of the arguments that informed the development of the project are outlined in our article Algorithmic brand culture: participatory labour, machine learning and branding on social media (open access here).
