To analyse the content of large set of images, image recognition systems allow transforming them to set of labels. However, our research suggests that image recognition services may widely differ on the outcoming labels. Therefore, we have developed a measurement to assess the cross-service label agreement score applying word embeddings.
We have developed a workflow to identify how images circulate within Facebook groups and between Facebook groups and the wider internet. Please contact Matti to get a dump of the code file while we polish the code for publication.
A Marxist language model is a specialized AI model fine-tuned on the texts of Karl Marx to incorporate a Marxist worldview into its output, focusing on class struggle, labor, and critiques of capitalism. It differs from baseline models by analyzing society and data through a materialist, value-laden perspective.
Our group has background in human-computer interaction, and this is clearly visible on our methodological choices. We build on approaches where design and prototypes are used together with social science know-how and approaches.