ASOK15636U Digital Methods: From Ethnography to Supervised Machine Learning

It is widely recognized that growth in digital data formats enable new relations between quantitative and qualitative methods of inquiry and analysis, thus posing questions as to how such new complementarities are best exploited for social-scientific and practical purposes.

This course will run through a number of ways in which qualitative and quantitative modes of inquiry complement one another. They will compose a research strategy that ensures both qualitative methods strength in valid interpretation of the meaning and consequence of social actions and quantitative methods strength in generalization and structural analysis. 

The course will structured around a mini research project from data collection to analysis. In the data collection part we will combine quantitative sampling theory and qualitative case selection theory to ensure a good sample. For exploring the data we will discuss and use a set of methods meant to map out ones data and display the varying densities and differences in ones data(network analysis, clustering, multidimensional scaling).

These maps work as navigational device to ensure that one digital ethnography gets hold of the relevant variation in ones data. The digital ethnography entails a practice of participant observation in digital spaces (e.g. a Facebook group, a company intranet board) for the purpose of learning the details of the values and motivations characteristic of social interaction in this setting.

The digital ethnography provides the interpretative grounding which insures that the categories and social processes, that in the later stage will be quantitatively assessed, have a hold in the meaning-making practices of the actors themselves. In order to explore, generalize and test large scale patterns we need to translate our ethnographic description into something quantifiable.

We will use supervised machine learning in order to scale our qualitative categorization/coding. The student will learn the pro's and con's of various strategies regarding ones sampling, optimization strategies, the uses of unsupervised methods as input.

Importantly we will pay a lot of attention to biases detection and correction to ensure the reliability and validity of our supervised machine learning model. This categorized dataset can then be used of more or less simple quantitative analysis which together with the digital ethnography will make the mini analysis.