Factitious disorder and its online variant Munchausen by Internet: understanding motivation and its impact on online users to develop a detection method
University College Cork
The overarching aim of the research in this thesis was to develop a method of detecting Munchausen by Internet (MbI) and garner an understanding of the dynamics of online communities faced with MbI. Ground work studies were required to learn more about the disorder, to decide exactly what method of detection would be most appropriate. This involved a review of the existing literature available on MbI (paper 1; Munchausen by Internet). It also involved conducting two studies which focused on experiences from the perspective of those with Factitious Disorder (FD) (paper 2; When the lie is the truth: Grounded theory analysis of an online support group for factitious disorder) and MbI from the perspective of victims (paper 3; Claiming someone else’s pain: A Grounded theory analysis of online community user's experiences of Munchausen by Internet). Both these studies were necessary as FD and its online variant MbI are some of the most poorly understood and under researched pathologies. This is primarily because of the difficulty in obtaining and retaining participants who have experience of the disorder. Therefore, what was previously known about the disorder was largely speculative. The research presented in this thesis overcame the issue of recruitment and retentions of participants, by analysing the first-hand accounts written online by those who have experience of the disorders. The information obtained from the two groundwork studies was used in the third study to decide on and develop an appropriate method of detecting MbI and for interpreting the discriminate attributes (paper 4; Detecting Munchausen by Internet: Development of a Text Classifier through Machine Learning). Beyond applying the findings of these studies to the development of the classifier, they also made new theoretical contributions to the existing literature on FD and MbI. The first two studies provide the very first large-scale studies of FD and MbI, using first-hand accounts from those it directly affects rather than observations that are speculative. Grounded Theory was used to analyse the text as it does not require an a priori theoretical framework but allows the data to build the theoretical framework itself, resulting in more innovative findings. The findings offer a new perspective of FD, one which contrasts with traditional theories and indicates that FD may be closely aligned with addiction. The second study examined the dynamics within an online community faced with MbI. The primary findings were that MbI users were targeting ‘ideal victim’ persona which offered protection from suspicion and increased the level of attention and sympathy they could receive. The presence or possible presence of MbI also resulted in members of online communities using strategies to avoid false accusations or being duped. These strategies had the unfortunate consequence of potentially eroding the therapeutic benefits of online communities, in particular personal empowerment, by restricting opportunities to confer normality and cultivate interpersonal support. In addition, the methods used by online community members and their moderators to detect MbI were uncovered. It typically involved high-level deception cues which raised suspicions and the checking authoritative references to confirm or refute these suspicions. The findings from study one and two, as well as the literature review from paper one, offered no overt cues which could be consistently attributed to MbI and offered no support for the feasibility of psychometric testing to detect MBI. Therefore, it was decided that covert deception required a covert method of detection. To this end the SLP (Social Language Processing) framework, which integrates psychology and computer science, was applied to develop a text classifier through machine learning algorithms. This covert method has already been successfully used to detect written deception online. Two text classifiers were developed in study three using Linguistic Inquiry Word Count (LIWC2105) dimensions and n-grams obtained from a bag-of words model, with respective prediction accuracies of 81.11% and 81.67%. These classifiers added a practical application value to the research conducted in this thesis, by producing a method of detecting MbI that can be used by moderators and as a vetting and investigative tool for internet mediated researchers. There were also theoretical contributions obtained from study three. Some of the discriminate attributes used by the classifiers appeared to be unique to Munchausen’s and were associated with the motivation for the behaviour, which supports the growing move towards domain specificity when interpreting Linguistic Based Cues (LBC) of deception. The remaining LBC’s of deception concurred with established deception theory, particularly reduction of cognitive complexity. Overall the research described in this thesis has made new contributions to the existing theories surrounding Factitious Disorder (FD), MbI and Linguistic Based Cues (LBC’s) of deception. It also has a practical application value by creating a classifier which differentiates between text written by genuine people and those exhibiting Munchausen’s.
Munchausen by Internet , Cyberpsychology , Factitious disorder , Machine learning , Text classification
Lawlor, A. 2018. Factitious disorder and its online variant Munchausen by Internet: understanding motivation and its impact on online users to develop a detection method. PhD Thesis, University College Cork.