Decolonizing AI: Rethinking Big Data’s Relation to the Contemporary Subject

In the digital age, knowledge is produced and disseminated through algorithms and machines. However, these systems can lead to homogenization, where knowledge is standardized and conformed to a single style or format. This can have serious consequences, as it can obscure the original author’s perspective and value, leading to a loss of recognition and agency.
To understand homogenization, imagine a recipe book filled with similar-looking dishes. Each dish may look appetizing, but they all follow the same pattern and style. Similarly, when knowledge is standardized and conformed to a single format, it loses its unique flavor and character, becoming indistinguishable from one another.
One way homogenization occurs is through incentives that encourage workers to make their content more easily extractable by machines. This can lead to a uniform style of writing that favors what the machine can best extract, rather than highlighting the worker’s unique perspective and expertise.
Another way homogenization can occur is through recommendation systems that suggest the same information to everyone searching for a particular subject. This can create a sociotechnical feedback loop where knowledge is constantly being homogenized towards a single expression of it.
To reduce the risks of homogenization, it’s essential to design systems that preserve context and value multi-modal knowledge types. Workers should be able to record knowledge in flexible structures and varied styles, and systems should support the use of multiple proxies to measure expertise. By doing so, we can create a more diverse and inclusive digital landscape where knowledge is valued and recognized for its unique contributions.
In conclusion, homogenization is a significant risk in algorithmic knowledge production. It can lead to a loss of recognition and agency for workers, as well as a loss of diversity and uniqueness in the knowledge produced. By designing systems that prioritize context and multi-modal knowledge types, we can create a more equitable and inclusive digital landscape where knowledge is valued and recognized for its true worth.