Counter-masquerading : a logicist-AI approach to interventionist strategies

Abstract

This dissertation introduces a novel framework for addressing the challenge of authorship attribution in the social-media sphere; the framework specifically seeks to contribute to attempts to achieve effective counter-masquerading. This framework, and the methods upon which it is based, touch on classical approaches to authorship attribution to detect masquerading, and also marks the inception of an entirely new sub-field of counter-masquerading: viz., interventionist strategies therein. The interventionist approach is an interactive approach in which the user suspected to be faking their identity is asked directly or indirectly to interact with designated questions or texts to gauge their reaction, and thereby to perhaps give rise to "cognitive inconsistencies,'' the presence of which aid detection of online impersonation. A key aspect of our method is (to our knowledge) the first use of counteridenticals. This type of conditional sentence/formula is a proper subclass of counterfactuals, and involves comparison of two identities in its antecedent. A pertinent example of such a conditional in the literary sphere would be : ``If Moncrieff were Proust, "Remembrance of Things Past" would have less Edwardian intonations.'' What we dub 'deep' counteridenticals compare incompatible identities within the purview of a "deep" pragmatic interpretation (this of course to be fully explained in the sequel).

In addition, we use a new expressive framework for the modeling of emotion to delve deeper into the behavior, desire, and emotional triggers in individuals who project their online presence. It is hard for a masquerader to mimic the psyche of an individual he/she seeks to "be,'' and the interventionist approach tries enable deeper insights into masqueraders than what these deceiving agents try to project virtually.

From times immemorial, masquerading has been a problem that has taken many shapes and forms. From physically impersonating a person to writing on behalf of someone, the degree and sophistication of such impersonation has varied considerably. Detection of masquerading has of course been with us as long as masquerading itself; and such detection is often a very difficult challenge. With the advent of social media, many new and often deleterious ways of masquerading have arrived on the scene. The influence of social media can, we all know, sometime shape opinions; the shift of modern life toward (at least in part) a virtual one has given rise to a plethora of new problems about online authorship, which is commonly termed the cybersecurity problem of masquerading. Simple stylometry, which long included finding penman strokes, grammatical patterns, and finding types of words used has now been "upgraded'' to authorship attribution, using 21st-century natural-language processing techniques (NLP) on large datasets. Counter-masquerading is still an embryonic, fast-developing field with new techniques coming out, it seems, every week, and because counter-masquerading is difficult, there is plenty of room for improvement.