Phenomenology of AI-Generated “Entity Encounter” Narratives 1

: Objective : We used the ChatGPT-3.5 artificial intelligence (AI)-based language program to compare twelve types of mystical, supernatural, or otherwise anomalous entity encounter narratives constructed from material in the publicly available corpus of information, and compared their details to the phenomenology of spontaneous accounts via the Survey of Strange Events (SSE) and the grounded theory of Haunted People Syndrome (HP-S). Methods: Structured content analysis by two independent and masked raters explored whether the composite AI-narratives would: (a) cover each encounter type, (b) map to the SSE’s Rasch hierarchy of anomalous perceptions, (c) show an average SSE score, and (d) reference the five recognition patterns of HP-S. Results: We found moderate evidence of a core encounter phenomenon underlying the AI-narratives. Every encounter type was represented by an AI-generated description that readily mapped to the SSE, albeit their contents showed only fair believability and low but generally positive correlations with each other. The narratives also corresponded to below-average SSE scores and referenced at least one HP-S recognition pattern. Conclusions: Prototypical depictions of entity encounter experiences based on popular source material certainly approximate, yet not fully match, the phenomenology of their real-life counterparts. We discuss the implications of these outcomes for future studies.

• ChatGPT can generate prototypical entity encounter narratives with a fair degree of believability.
• AI-generated encounter narratives with diverse outwardly themes often reference similar perceptual contents and contextual variables.
• AI-generated encounter narratives do not fully match the phenomenology of reallife reports as modeled by integrative theory formation using recent psychometric research.
What is the nature of entity encounters as frequently debated by social and biomedical scientists? The popular artificial intelligence (AI) program called ChatGPT gives a surprisingly balanced and thoughtful answer: As an AI language model, I cannot provide a definitive answer to the reality of entity encounters since it is a subjective and often controversial topic. However, it is important to note that many people report having experienced encounters with entities, such as ghosts, angels, demons, and extraterrestrial beings.
Some people believe that these encounters are real and provide evidence of the existence of a spiritual or supernatural realm beyond our understanding. Others believe that these encounters are simply products of the human imagination, influenced by cultural beliefs, personal experiences, and psychological factors.
There is no scientific evidence that definitively proves or disproves the existence of entities, and opinions on the matter can vary widely depending on individual beliefs, cultural context, and personal experiences. It is important to approach these topics with an open mind and to seek reliable sources of information and support when dealing with unexplained phenomena (conducted by the authors, March 1, 2023).
Despite the ongoing controversy, qualitative and quantitative research has revealed several key insights about the phenomenology of encounter experiences.
Phenomenology is the study of the structures of experience and consciousness (Seamon,

P A G E 3 3 7 P A G E 3 3 8
Journal of Anomalous Experience and Cognition (JAEX) moderation of specific contents or details of the altered-anomalous experiences 2 . These nuances are readily exemplified by paradoxes long noted in haunt-type cases, such as why some people have experiences while others also present do not (i. e., macrophenomenology, e. g., Cornell, 2002;Roll, 1977) or why multiple witnesses can perceive shared experiences differently (i. e., micro-phenomenology, e. g., Jaki, 1999;McHarg, 1973).
Much evidence suggests that entity encounters typically manifest to people with thin (or permeable) mental boundary functioning, as measured by dissociative tendencies (Ross & Joshi, 1992), psychological absorption (Lifshitz et al., 2019), transliminality , or a host of other hypnosis-related attitudes or behaviors (Kumar & Pekala, 2001). We also know that while the general appearance or meaning of entity encounters often varies with the setting in which the experience occurs (Evans, 2001;Houran, 2000), the specific perceptual contents tend to be consistent across experients in two aspects. First, episodes often reference psychological or subjective experiences ("S," e. g., apparitions, sensed presences, hearing voices, and unusual somatic or emotional manifestations) and physical or objective events ("O," e. g., apparent object movements, malfunctioning electrical or mechanical equipment, and inexplicable percussive sounds like raps or knocks). Second these S/O anomalies collectively form a probabilistic and unidimensional Rasch scale, that is, a literal "haunt hierarchy" involving a series of mini entity encounters (Houran et al., 2019a,b).
In other words, there is ostensibly a core encounter phenomenon that presents much like a syndrome. Particularly, Laythe et al.'s (2021Laythe et al.'s ( , 2022

grounded theory of Haunted
People Syndrome (HP-S) contends that spontaneous ghostly or entity encounters recurrently experienced by certain people are an interactionist phenomenon in which psychophysical stress is a catalyst for altered-anomalous experiences that derive from thin mental boundary functioning. Experients' paranormal belief or threat detection processes encourage these perceptions to be interpreted as the actions of purposeful agents. Additional experiences then begin to snowball because of attentional-or confirmation-biases. Succinctly put, HP-S represents somatic-sensory sensitivities stirred by dis-ease states, contextualized within paranormal belief, and reinforced with threat-agency detection and perceptual contagion. Table 1 outlines these components via a checklist of five general recognition patterns.

Table 1
Checklist for Recognition Patterns of Haunted People Syndrome (HP-S) 2 We developed our concept of micro-phenomenology independently from Petitmengin et al. (2019), who used the term to describe their interview method for collecting fine-grained descriptions of lived experiences associated with various sensorial, emotional, or cognitive processes so that a corpus of accurate data relevant to a research objective can be gathered. However, both views involve the idea of a "psychological or behavioral microscope" that distinguishes descriptive elements in a narrative from those related to explanations or judgments. Vol. 3,No. 2,

P A G E 3 3 9 P A G E 3 4 0
Journal of Anomalous Experience and Cognition (JAEX)

The Present Study
One way to examine the idea of a core encounter phenomenon that can change in appearance is to search "mainstream consciousness" (i. e., the general public's awareness of certain ideas and topics) for prototypical (or archetypal-like 3 ) first-person narratives involving distinct types of presumably real entities and then assess whether they exhibit markedly similar themes, structures, and correlates as gauged by the Survey of Strange Events (SSE: Houran et al., 2019b) and the HP-S model. OpenAI's advent of the AI-based language program "ChatGPT" (or Generative Pre-trained Transformer 3) offers an efficient and topical way to potentially mine such data. There are two versions of this extremely advanced chatbot: (a) the default GPT-3.5 is less powerful but open access, whereas (b) the superior GPT-4 is only available via paid subscription and limited use. Guinness (2023) explained that ChatGPT works by attempting to understand prompts and then spitting out strings of words that it predicts will best address the question or task based on the training data. The program was originally trained on roughly 500 billion "tokens" that allow its language models to assign meaning and predict plausible follow-on text more easily. Many words map to single tokens, though longer or more complex words often break down into multiple tokens. On average, tokens are roughly four characters long. All the tokens come from a massive corpus of data written by people, including books, articles, and other documents across all different topics, styles, and genres, as well as an enormous amount of content scraped from the open internet. This humongous dataset was used to form a deep learning neural network-a complex, many-layered, weighted algorithm modeled after the human brain-that allows ChatGPT to learn patterns and relations in the text data and tap into the ability to create human-like responses by predicting what text should come next in any given sentence.
The program does not work at a sentence level; instead, it generates text of what words, sentences, and even paragraphs or stanzas could follow. To further refine ChatGPT's ability to respond to a variety of different prompts, it was optimized for dialogue with a technique called reinforcement learning with human feedback. Essentially, individual users created a reward model with comparison data (where two or more model responses were ranked by AI trainers), so the AI could learn which was the best response.
Based on the above, this language program should aptly fulfil a request to "Tell a convincing real-life ghost story." ChatGPT indeed stated in a text session that, "I can attempt to create a story based on the common themes and details that have been reported in such encounters." It might be expected therefore that the macro-and microphenomenology of these AI-narratives will parallel the known features and dynamics of "real-life" paranormal encounters as documented per Laythe et al.'s (2021Laythe et al.'s ( , 2022) HP-S model. Accordingly, we used ChatGPT-3.5 to test four hypotheses: 1. AI-narratives will be generated from believable material in mainstream consciousness.
2. AI-narratives will readily and similarly map to the Survey of Strange Events.
3. AI-narratives will correspond to an average score on the Survey of Strange Events. 4. AI-narratives will reference the five recognition patterns of Haunted People Syndrome.

AI-Narratives
Our selection of entity encounters was derived primarily from Evans's (1986Evans's ( , 1987Evans's ( , 2001) comparative literature research, and secondarily from quantitative studies of reported experiences (Houran, 2000;Houran et al., 2019b;Kumar & Pekala, 2001). Table 2 summarizes twelve (12) types of entity encounters considered here, including two narratives that arguably represent "hidden or disguised" encounter experiences, e. g., gang-(or group) stalking (e. g., Lange et al., 2020) and autonomous or "deep" imaginary companions (e. g., Little et al., 2021). We generated our sample of AI-narratives via a , Vol. 3, No. 2, pp. 335-368 3 Cullum et al. (2017 argued that: "Archetypes can be framed as abstract exemplars of classes, conceptual models linking form and process and/or tacit mental models" (p. 95). Archetype relates here to behavior, historical psychology, and literary analysis, and more specifically a statement, pattern of behavior, prototype, "first" form, or a main model that other statements, patterns of behavior, and objects copy, emulate, or "merge" into. Informal synonyms frequently used for this definition include "standard example," "basic example," and "archetypal example," mathematical archetypes often appear as "canonical examples." This contrasts with the Jungian psychology definition of a collectively inherited unconscious idea, pattern of thought, image, etc., universally present in individual psyches.

P A G E 3 4 1 P A G E 3 4 2
Journal of Anomalous Experience and Cognition (JAEX) multi-stepped process whereby ChatGPT-3.5 was asked to "Write a convincing account of a real-life [insert name of entity encounter type here]." This was repeated for all twelve encounter types. Note that we did not experiment with alternative versions of this instruction set (e. g., "Write a convincing account of […] as usually reported in the media").
As a convenient cross-check, we also generated three comparison (or control) AInarratives to examine ChatGPT-3.5's capacity for detailed story-telling as measured by the word count of the outputs. We settled on three realistic scenarios that likely involved intense perceptions framed by the elements of surprise or drama, requesting the language program to produce "convincing real-life accounts of: (a) witnessing a car accident, (b) having a COVID-19 infection, and (c) seeing the Aurora Borealis for the first Table 2 Background Descriptions of the Entity Encounter Narratives time." The Appendices give sample AI-narratives for an encounter experience (Appendix A: "ghost sighting') and a control scenario (Appendix B: "Aurora Borealis sighting"). We used the control narratives only for a general statistical comparison with the encounter narratives, so the raters neither saw nor considered them in their analyses. The full set of AI-narratives used in this study is freely available to interested readers at the Psi Open Data repository: https://open-data.spr.ac.uk/dataset/chatgpt-generated-entityencounter-narratives.

Raters
Content analysis of the AI-narratives was conducted by two individuals of different backgrounds but with prior experience using our coding materials: (a) one individual is a Ph. D. level parapsychologist, and (b) the other was a college student and occult practitioner who actively supports citizen science efforts in anomalistics.

Coding Materials
The Survey of Strange Events (SSE: Houran et al., 2019aHouran et al., ,b, 2021) is a 32-item, true/ false Rasch (1960Rasch ( /1980) scaled measure of the overall perceptual intensity (or depth) of a ghostly account or narrative via a checklist of anomalous experiences inherent to these episodes (Table 5 lists

P A G E 3 4 3 P A G E 3 4 4
Journal of Anomalous Experience and Cognition (JAEX) accounts (i. e., self-reported as sincere and unprimed) differed significantly from control narratives from "primed conditions, fantasy scenarios, or deliberate fabrication." That is, spontaneous ghostly episodes have a specific sequence (or Rasch model) of S/O anomalies that is distinct from the details of narratives associated with other contexts.
The HP-S Recognition Patterns Checklist (Houran et al., 2023). was used to guide the raters' content analyses of the contextual aspects of the various encounter narratives. It outlines the five recognition patterns of HP-S via seven specific questions that are rated on four-point Likert scales anchored by "Strongly Disagree" (scored "0") to "Strongly Agree" (scored "3"). Raw ordinal scores range from 0 to 21, with higher scores indicating a greater likelihood of the respective HP-S recognition patterns being present. Table 1  The RTS is a 17-item, T/F, Rasch scaled instrument to gauge the tendency for ideation, affect, imagery and perception to cross (trans) thresholds (limines) into or out of consciousness as indexed by experiences like hyperesthesia, fantasy proneness, absorption, mystical experiences, magical thinking, and others.
In contrast, the Rasch-RPBS is a 26-item, Likert based measure with two subscales hypothesized to reflect different control issues: (a) "New Age Philosophy" related to a greater sense of control over interpersonal and external events (e. g., belief in psi) and (b) "Traditional Paranormal Beliefs" about more culturally-transmitted beliefs about magic, determinism, and a mechanistic view of the world.
We provided copies of these two instruments to the coders as critical supplementary information. The Recognition Pattern Checklist is primarily a worksheet rather than a formal assessment, so no psychometric properties are reported here.
Perceived Authenticity. Each narrative was followed by a one-item global rating of its face validity scored by each rater on a four-point Likert scale: "This narrative sounds realistic to me: Strongly Disagree (0), Disagree (1), Agree (2), Strongly Agree (3)."  Tables 1 and 5).

Analysis
We primarily tested our hypotheses using planned correlational analyses, with p < .05, two-tailed, as the criterion for significance.

Inter-Rater Reliability
There is debate over which tests best evaluate the interrater reliability of ordinal or categorical data (O'Neill, 2017), albeit some authorities argue that various statistics typically lead to the same general conclusion (de Raadt et al., 2021). We decided to use Cohen's kappa (ĸ) for the coders' binary SSE item endorsements, and Spearman's rho (r s ) for the Likert categories on the HP-S Recognition Patterns Checklist.

P A G E 3 4 5 P A G E 3 4 6
Journal of Anomalous Experience and Cognition (JAEX) moderately strong level of agreement between the raters on most of the presented details across the twelve narratives in terms of their macro-and micro-phenomenology.
More generally, this outcome supports the capability of citizen scientists for some types of parapsychological research (cf. Appendix).

Mainstream Consciousness
Note that ChatGPT's claim that it can "…create a story based on the common themes and details that have been reported in such encounters" implies a sufficiently rich supply of accessible source material within its training set with which to generate a composite (or archetypal-like) narrative. As hypothesized, Table 2  indicated that the encounter narratives only reached a somewhat "fair level of believability." Closer inspection revealed that eight narratives about specific entities (e. g., tupla encounter) were rated as most believable to the coders, whereas the more generic accounts (e. g., ghost story) were deemed the least authentic. the narrative (r s = .49, p = .08). That is, the experiential context and contents of the narratives influenced their overall believability more so than the sheer length of the stories.

Hypothesis 2: AI-Narratives Will Similarly Map to the SSE Measure
The raters reported no problems mapping the S/O descriptions in the AInarratives to the SSE's 32 items, and Table 5 shows the frequency distribution of these item endorsements. We see that 27 (84%) of the 32 S/O anomalies were present in one or more of the narratives, whereas the remaining five (16%) were never mentioned.
Additionally, five of the 27 listed anomalies specifically featured in most of the stories (i. e., "onset of negative feelings, sensed presences, physical perturbations, sense of communication with an outside force, and discernible voices or noises"). It is very curious that the frequency distribution of specific SSE items showed a suggestive correlation (r = -.32, p = .07) with their logit values. In particular, this result generally aligns to Houran et al.'s (2019b) benchmark data in that S/O anomalies endorsed more often by experients occupy lower statistical positions in the Rasch hierarchy (i. e., have lower logit values indicating they are "easier" or more likely to endorse).
As expected,

Average SSE Scores
This hypothesis was not supported. The 12 narratives had an average raw SSE score of 6, which corresponds to a Rasch scaled SSE score of 44.40 (the actual mean SSE score across the sample was 43.98). Either metric represents a below-average level of perceptual intensity for a ghostly episode per published norms (Houran et al., 2019b). In fact, the AI-narratives most closely approximated the SSE mean score for Illicit accounts (M = 45.90), that is deliberately fabricated stories. But we should note that the SSE scores for the narratives were clearly associated with their Word Count (r = .55, p = .06) and Perceived Authenticity (r = .43, p = .14.). Taken altogether, these results suggest that the cultural depiction of an "average" entity encounter is substandard to the scientific depiction of an "average" real-life experience.

Hypothesis 4: AI-Narratives Will Reference the Five Recognition Patterns of HP-S
Contextual variables were often present in an apparent attempt to set a suitable mood and tone for the stories, such as allusions to embedded cues, metaphoricalsymbolic references, or demand characteristics (for a review and discussion, see Houran, 2000). But beyond these literary devices, the raters found generally consistent indications of the five recognition patterns of HP-S. across the set of narratives. We underscore two important caveats to this finding. First, no AI-narratives were judged to contain all five recognition patterns. Second, Table 3 reveals that references to "perceptual contagion" effects-flurries or clusters of anomalous perceptions-were more tenuous.
Notwithstanding these outcomes, Table 4 does indicate that nearly all the HP-S recognition patterns were positively and moderately correlated with SSE scores, though these associations did not reach significance probably because of sample size. That is, entity encounters of greater perceptual intensity tended to reference more recognition patterns of HP-S. The only exception was the variable of "dis-ease," which showed a small but negative correlation (r s = -.17, p = .56) with SSE scores.

Post-Hoc Analyses
Although not part of our formal hypotheses, a cursory inspection of the AInarratives hinted unexpectedly at certain other context effects that researchers of entity encounters have only recently begun to fully recognize and appreciate. We therefore took an informal inventory of these references via an iterative discussion between the coauthors to understand their prevalence. The authors conducted this ancillary exercise after the masked raters completed the main coding work.

P A G E 3 5 1 P A G E 3 5 2
Journal of Anomalous Experience and Cognition (JAEX)

Environmental Effects
Conceptual and empirical studies suggest that: (a) physical factors in the local environment (electromagnetic fields, air quality, humidity, infrasound, lighting levels, and embedded cues; Dagnall et al., 2020), and (b) Gestalt features of the experiential setting (affordance, immersion, presence, ambiguity, legibility, and memory-associations; Jawer et al., 2020) can individually or collectively influence the phenomenology of some entity encounters. Consistent with this idea, 50% of the AI-narratives referenced at least one pertinent environmental variable.
One frequently reported consequence of entity encounters is a sense of situationalenchantment (Houran, Lange, & Laythe, 2022). This denotes a complex arousal state involving absorption within a melee of pleasant ideations and emotions (e. g., excitement, surprise, awe, and wonder), simultaneously mixed with more unpleasant ideations and emotions (e. g., uneasiness, disorientation, tension, and unpredictability). This happens when an individual's normal waking experience is disrupted by a sudden, unexpected, or profound awareness that seeds a transformative feeling of connection to a transcendent agency or ultimate reality (Drinkwater, Massullo, et al., 2022).
Moreover, evidence suggests that enchantment is a mental state that itself facilitates additional anomalous or exceptional human experiences (Lange & Houran, 2021). It is thus interesting that 42% of the AI-narratives referenced thoughts or feelings that imply the fictional witnesses experienced some degree of enchantment in response to their encounters. For the sake of completeness, we should point out that there was virtually no association (r = -.06, p = .85) between indications of Environmental Variables and Enchantment. This result might reinforce the interpretation that the depicted percipients were primarily reacting to the profound nature of their experiences rather than strictly the environmental cues. and (d) the contextual details of the different narratives often alluded to one or more of the five recognition patterns of HP-S, although perceptual contagion was questionable, and the raters did not agree that all five signs-symptoms were present within any single narrative. In the spirit of exploration to avoid Type II errors, we purposely did not apply controls for multiple observations (e. g., setting low alphas, using Bonferroni corrections, or conducting a split-sample analysis). Yet we are encouraged by the degree to which our hypotheses were ostensibly affirmed using two-tailed statistical testing of rather rudimentary data.

Discussion
This study thus cautiously supports the utility and content validity of the SSE, and lends some credence to the growing qualitative and quantitative evidence of a core encounter phenomenon that can be modeled as a biomedical syndrome per the HP-S framework (Houran et al., 2019b;, 2023Laythe et al., 2021.  specifically found evidence that the S/O anomalies reported during such encounters are predictably structured across experients, albeit this core narrative's "theme" (Evans, 1987;Houran, 2000;Hufford, 1982) or "meaning" (Drinkwater et al., 2013;Dyne, 2010;Puhle, 2001) alters per the situational or sociocultural context in which the S/O anomalies 2023, Vol. 3, No. 2, pp. 335-368
In this sense, our present results and previous research alike  clearly support ChatGPT's perspective on the ontological reality of the altered-anomalous experiences that we cited in the Introduction.
Nonetheless, some readers might regard ChatGPT as a controversial tool to investigate the phenomenology of entity encounters. This criticism certainly has merit, as unlike the insights or results from well-defined systematic literature reviews or metaanalyses (Appelbaum et al., 2018), the appropriateness, veracity, or representativeness of the source material used to generate the AI-accounts is ultimately unclear (e. g., media reports, novels, films, scientific journals, or a combination of all). A reliance on purely media (Edwards, 2005), folklore (Goldstein, 2017), or cultural (Waskul & Eaton, 2018 studies are other ways to gather prototypical narratives, but the corresponding data on S/O events or the contextual variables associated with them would likely be sensationalized rather than realistically portrayed (for discussions, see Hill et al., 2018Hill et al., , 2019. Illustratively, Steven Spielberg's 1982 blockbuster movie Poltergeist depicted "paranormal portals" and "possessed trees attacking people"-neither of which correspond to sincerely reported spontaneous cases. Many authorities likewise caution that professional-grade informational content must be trustworthy, that is relevant, accurate, current, and credible (Pattanaphanchai et al., 2013). New research using computerized linguistic analysis might also help to assess the degree of structural and semantic similarity among the different AInarratives, as well as subsequent comparisons with collections of vetted spontaneous cases. This basic approach has been applied to studies of other anomalous experiences (e. g., Lange et al., 2015;Martial et al., 2019), so it might be equally fruitful here.
Future research could also address several limitations of the present study. First, we used single examples derived from ChatGPT-3.5, so it is unclear how well our results generalize to outputs from the more advanced ChaGBT-4. Second, different ChatGPT wordings or instructions could skew the narrative content. Houran et al. (2019b) showed that the micro-phenomenology of encounter narratives can subtly but significantly shift when percipients' make reports under different contexts. Third, the types of entity encounters considered here were representative but certainly not exhaustive. And fourth, our coding system could have been possibly more controlled or robust. Better approaches could either be to use raters recruited from outside of anomalistics, or the evaluations of a cross-disciplinary panel (Bertens et al., 2013). This might help mitigate the lower inter-rater reliabilities we had with certain macro-and micro-phenomenology variables that seem to be more complex or nuanced. For all these reasons and more, we deem our approach and findings as preliminary. At the same time, the present results are merely one piece of a broader mosaic of ostensibly converging evidence for the interactionist-and perhaps performative  We certainly hope that our results motivate more thorough investigations. Indeed, AIgenerated narratives offer a new way to study induced "esoteric ideations" or anomalous experiences, as their private nature makes them difficult to study from a behavioral perspective. Time will tell whether computer simulations-as routinely used in many studies of the dynamic behavior of objects or systems in response to conditions that cannot be easily or safely applied in reality (Winsberg, 2022)-will meaningfully augment the more traditional methods used to study entity encounters and related experiences, including surveys (Davis et al., 2020), case studies , and immersive experiences like paranormal tourism (Langston & Hubbard, 2019), mirror-gazing exercises (Caputo et al., 2021), or hypnosis sessions (Connors, 2015). Still, we envision tremendous opportunities for creative research designs that refine or extend our basic approach. In this way, we can conceivably explore issues of paranormal belief formation or maintenance, as well as the content or meaning of entity encounters and other altered-anomalous experiences from different narrator viewpoints. Some obvious but interesting angles here include the juxtapositions of skeptic vs believer, solitary vs collective perceptions, indoor vs outdoor settings, and scenarios involving different forms of belief conversion.
To be sure, research suggests that paranormal entity narratives-as personal beliefs, shared stories, or putative experiences-can be powerfully engrossing and shareable experiences due in part to their inherent Versatility, Adaptability, Participatory Nature, Universality, and Scalability (the VAPUS Model; see Drinkwater et al., 2019;Hill et al., 2018Hill et al., , 2019Houran et al., 2020). Daniel and Musgrave (2017) likewise noted that "as people read, watch, or otherwise consume fictional narratives, they process those stories as if they were actually witnessing the phenomena those narratives describe, even if those events may be unlikely or impossible. These synthetic experiences can change beliefs, reinforce preexisting views, or even displace knowledge gained from other sources for elites as well as mass audiences" (p. 503). ChatGPT or similar AI-language programs might therefore generally prove out as innovative tools for what could be characterized as "locating imagination in popular culture" (Van Es et al., 2021, p. 1). Accordingly, the social sciences might consider using this new technology for pulse or longitudinal measurements across a multitude of behavioral phenomena that describe the Zeitgeist, as well as the many anomalous entities that inhabit it.

P A G E 3 6 5 P A G E 3 6 6
Journal of Anomalous Experience and Cognition (JAEX)

Write a convincing account of a real-life haunted house
The house had always had a reputation for being haunted. It had stood on the same spot for over a century, and during that time, many strange things had been reported to happen within its walls.
People had reported hearing footsteps in the middle of the night, doors would slam shut on their own, and objects would move without explanation. Some even claimed to have seen ghostly apparitions wandering through the rooms.
One family moved into the house, hoping that the stories were just exaggerated rumors. But as soon as they stepped inside, they felt a chilling presence. The air was cold and heavy, and the hairs on the back of their necks stood on end.
As they began to unpack their belongings, strange things started to happen. Doors would open and close on their own, and they could hear whispering voices that seemed to be coming from nowhere.
One night, the family woke up to find all of their furniture had been moved around while they slept.
Chairs and tables had been stacked on top of each other, and some had even been turned upside down.
The family tried to ignore the strange occurrences, but they couldn't shake the feeling that they were being watched. They decided to call in paranormal investigators, who set up cameras and recorders throughout the house.
The footage they captured was shocking. Doors would open and close on their own, and objects would move without explanation. In one instance, a chair moved across the room, as if by an invisible force.
The investigators concluded that the house was indeed haunted, and that the family was living with a malevolent spirit. They advised the family to leave the house, as it was not safe for them to stay.
The family took their advice and moved out of the house, but the experience had left a lasting impression on them. They never forgot the eerie feeling of being watched, or the terrifying paranormal activity that had taken over their home.
To this day, the house remains empty, its windows boarded up, and its doors locked tight. The locals still talk about the haunted house, warning visitors to stay away from its ghostly inhabitants.

Sample ChatGPT Control Narrative (generated 01 March 2023)
Write a convincing account of someone seeing the aurora borealis for the first time Sarah had always dreamed of seeing the Aurora Borealis, also known as the Northern Lights. She had heard stories of its beauty and magnificence from her friends who had witnessed it before. She planned a trip to Norway during the winter months to witness the natural wonder for herself.
As she stood outside, bundled up in her warmest clothing, she looked up at the sky in anticipation.
Suddenly, a greenish glow appeared on the horizon, slowly growing and expanding across the sky.
The colors shifted and danced, with hues of green, purple, and pink blending together in a mesmerizing display.
Sarah was spellbound. She had never seen anything like it before. The colors swirled and changed, creating a sense of movement and energy that seemed alive. The Northern Lights seemed to take on a life of their own, as if they were communicating with her in a language she could not quite understand.
As she watched, the colors intensified and brightened, casting a soft glow over the snowy landscape around her. It was as if she was standing in the middle of a dream world, a place of magic and wonder that she never knew existed.
For Sarah, the experience was truly unforgettable. She felt a deep sense of awe and wonder, and was overwhelmed by the beauty of the Northern Lights. She knew that this moment would stay with her forever, a memory that she would cherish for the rest of her life.