The impact of paratext on readers of generative literature: Human evaluation of generated text

Abstract

Generative literature has been emerging as a creative form since before the digital era, with one of its defining characteristics being the particular circumstances of its production. Since 2013, the National Novel Generation Month (NaNoGenMo) challenge has been one of the most prolific organized events and archives of generative literature in the form of computer generated novels. Authors present their generated novel projects as a collection of digital files and components which have different functions: the generated text, the code used to generate it, the author’s notes and creative intentions, the input data used by the code, and so forth. While generated novels and other forms of generative literature have been researched to a limited extent, little has been investigated about readers and their experience, especially beyond an implied reader perspective as theorized through an academic lens. With the presence of generative qualities and processes becoming more apparent and increasingly present in today’s tools, in (creative) media, and in the sociotechnical fabric more generally, it is becoming increasingly relevant to research actual readers’ understanding and reception of generated text. However, as both an emerging creative and technical form with many different components, it's unclear which components are central to impacting readers’ understanding and reception of a generated novel.
The impact of paratext on the interpretation and reception of media works has long been accepted and theorized in literary studies. Paratext refers to the framing, supplementary, and periphery elements which can accompany a work. Paratext’s perceived ability to steer readers’ interpretations about a piece was first detailed by structural literary theorist Gérard Genette’s established but highly contested conceptualization of paratext. Genette theorizes how paratext functions in the context of books and their consumption, and his conceptualization has since been frequently used to describe new media works and their audiences, although not without challenge by new media forms. However, within this literary and media space there is little in the way of participant-based research which tests assumptions about how paratext functions to impact different readers’ interpretation and reception of a work. Nevertheless, this research project demonstrates that Genette’s paratextual conceptualization lends itself to being easily operationalized such that the impact of different elements on reading reception can be measured and unpacked using quantitative and qualitative methods.
Focusing on two distinct NaNoGenMo generated novels, this research project conceptualizes the components of a generated novel as paratextual elements in order to test the ways in which paratext impacts potential readers’ interpretation and reception of a generated novel. Reception is decomposed into different value dimensions such as literary, creative, and technical value, as well as understanding, interest, and enjoyment. The research project further investigates how reader’s computer programming skills or literary reading experience might interplay with the impact of the paratextual elements, as well as how cultural characteristics of a generated novel might shape interpretation and reception.
The research project is carried out as an explanatory sequential mixed-method design beginning with an online reading experiment where the generated novels are presented in one of three different paratextual conditions. Participant responses are recorded through a survey and a written review where participants give their personal opinions about the work they read. The survey results suggest that while the presence of the paratextual elements in the generated novels have a measurable impact on readers’ perceptions across virtually all the value dimensions, the presence of the generated text itself doesn’t significantly impact reader valuations. Further, the survey results also suggest that the generated novels’ perceived literary value may be largely resistant to being impacted by the paratext.
In addition to further studying paratext’s impact on reader interpretation and reception, this unexpected literary value result is further explored in the written review data using Qualitative Content Analysis, and in a second in-person reading group workshop study where a semi-structured group interview is run to discuss reader’s impressions of the same generated novels read in printed paperback book form – a physical paratext. The workshop study data is analyzed using Deductive Qualitative Analysis and develops a working conceptual model of a printed generated novel’s reading and interpretation process. Here, latent links to a generated novel’s technical aspects and a reader’s own creative framing are highlighted as primary themes. The analysis develops further and crystalizes into a novel contribution in the form of a minimalist theoretical refinement of Genette’s paratextual conceptualization. The findings also suggest that the readers’ choice to engage with the generated text ergodically and to interpret it through a creative framing appear to be factors which contribute to the readers perceiving their reading experience to be more enjoyable. Nevertheless, despite workshop discussion about the physical paratext, references to authorial paratext, and the generated novels’ cultural value, the literary value of generated novels appears to remain challenging for readers to fully accept.

More information

Identifiers

ORCID for Lesia Tkacz: orcid.org/0000-0002-5570-406X

ORCID for David Millard: orcid.org/0000-0002-7512-2710

ORCID for Megen de Bruin-Molé: orcid.org/0000-0003-4243-1995

Catalogue record

Export record