Browsing by Author "Koivusaari, Eeva Matleena"

Abstract Objectives. Despite the convincing evidence of efficacy of psychotherapy, its mechanisms of change in therapy are still unclear raising questions on what constitutes psychotherapy. In this work, the common and specific factors of cognitive behavioral therapy and psychodynamic psychotherapy were inspected by exploring psychotherapy manuals, which has not been previously studied in therapy research, using machine learning methods. The aim of this work was to gain understanding of the theoretical frames of different psychotherapy approaches and explore their distinguishability as well as semantic features. Methods. Textual psychotherapy manuals representing psychodynamic psychotherapy (n=4) and cognitive behavioral therapy (n=4) were analyzed using latent Dirichlet allocation topic modeling. The number of topics was pre-defined as 2, based on the number of therapy approaches. Additionally, a Bayesian method was used, defining the best fitting number of topics for the data. Logistic regression was performed for both solutions regarding the selection of the number of topics to observe the associations of textual topics with the therapy approaches. Two models were compared in their ability to correctly discriminate between the therapy approaches. Results and conclusions. The models consisting of 2 and 24 topics proved excellent in distinguishing the therapy approaches thus pointing towards the differentiable nature of these two approaches. The discriminative power of the model with 24 topics was based on few perfectly discriminative topics yet the model also included topics common for both therapy approaches. Based on this work, the approaches differed regarding relational and agency emphasis whereas shared expectations appeared as common factor.