Coda

The chapter concludes our exploration of the integration between computational social science (CSS) and sociology. We emphasize the necessity of incorporating sociological concepts of meaning and interpretation into CSS, which has traditionally focused on behavioral data analysis. Previous chapters demonstrated methods for embedding these concepts in agent-based modeling and digital data analysis. We also discuss the potential of CSS, particularly through machine learning, to transcend traditional deductive approaches in sociology. This advancement allows for the exploration of new theories beyond the constraints of existing sociological frameworks. However, we note a challenge in interpreting complex models generated by machine learning. To address this, we suggest the scientific regret minimization method, aiming to produce interpretable models with robust predictive capabilities. Furthermore, we explore the application of computational linguistic models, like topic models and word-embedding models, in understanding sociological meaning-making. These models can elucidate the processes of prediction, relationship, and learning in meaning-making. In summary, we argue that CSS's full potential in sociology can be realized by grounding its use in significant sociological ideas and theories. This approach not only leverages the analytical power of CSS but also contributes to the development of new sociological theories, thereby advancing the field.