Next Article in Journal
LFTD: Transformer-Enhanced Diffusion Model for Realistic Financial Time-Series Data Generation
Next Article in Special Issue
Epistemic Agency in the Age of Large Language Models: Design Principles for Knowledge-Building AI
Previous Article in Journal
Advances in Audio-Based Artificial Intelligence for Respiratory Health and Welfare Monitoring in Broiler Chickens
Previous Article in Special Issue
Pedagogical Transformation Using Large Language Models in a Cybersecurity Course
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
AI
Volume 7
Issue 2
10.3390/ai7020059
Due to scheduled maintenance work on our servers, there may be short service disruptions on this website between 11:00 and 12:00 CEST on March 28th.
ai-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Multisite Study of an Animated Cinematic Clinical Narrative for Anticoagulant Pharmacology Education

by
Amanda Lee
1,2,†,
Kyle DeWitt
1,2,†,
Meize Guo
3 and
Tyler Bland
2,*
1
University of Washington School of Medicine, Seattle, WA 98195, USA
2
WWAMI Medical Education Department, University of Idaho, Moscow, ID 83844, USA
3
The CS Everyone Center for Computer Science Education, University of Florida, Gainesville, FL 32611, USA
*
Author to whom correspondence should be addressed.
†
These authors contributed equally to this work.
AI 2026, 7(2), 59; https://doi.org/10.3390/ai7020059
Submission received: 2 January 2026 / Revised: 29 January 2026 / Accepted: 1 February 2026 / Published: 5 February 2026
(This article belongs to the Special Issue How Is AI Transforming Education?)
Browse Figures
Versions Notes

Abstract

Anticoagulant pharmacology is a cognitively demanding domain in undergraduate medical education, with persistent challenges in learner engagement, retention, and safe clinical application. Cinematic Clinical Narratives (CCNs) offer a theory-informed multimedia approach designed to support learning through narrative structure, visual mnemonics, and affective engagement. We conducted a multi-site quasi-experimental study within a six-week Cancer, Hormones, and Blood course across a distributed medical education program. First-year medical students received either a traditional case-based lecture or an animated CCN (Twilight: Breaking Clots) during a one-hour anticoagulant pharmacology session. Learning outcomes were assessed using pre- and posttests, learner engagement was measured with the Situational Interest Survey for Multimedia (SIS-M), and exploratory eye tracking with second-year medical students was used to assess visual attention to embedded mnemonics. Both instructional groups demonstrated significant learning gains, with fold-change analyses indicating greater relative improvement among students exposed to the CCN. The animated CCN elicited significantly higher triggered situational interest compared with non-animated cases (p = 0.019), while also being preferred by the majority of participants. Qualitative analysis revealed that learners perceived CCNs as particularly effective for initial encoding and memorization, while non-animated cases supported subsequent clinical application. Eye-tracking data demonstrated high visual uptake and sustained attention to key mnemonic elements. Together, these findings support expert-designed, genAI-assisted CCNs as a validated and complementary instructional approach in medical education.

1. Introduction

The rapid pace at which medical students are expected to learn large volumes of information has frequently been characterized as akin to “drinking from a firehose,” reflecting sustained cognitive overload during training [1,2]. Anticoagulant pharmacology, in particular, presents a set of foundational concepts that must be deeply understood to ensure safe and effective clinical practice [3,4]. Learners must integrate mechanisms of action, monitoring parameters, drug–drug interactions, and nuanced risk–benefit decisions, all of which place substantial demands on working memory and long-term retention [5,6]. Deficits in durable learning and memory in this domain may contribute to downstream medication errors, as anticoagulants remain among the drug classes most commonly associated with preventable adverse drug events in hospital settings [7,8]. While medical students are not primarily responsible for prescribing decisions during training, observational learning and early cognitive schemas formed during undergraduate medical education may shape future prescribing behaviors and error propagation [9,10,11].
Educational research has consistently demonstrated that traditional text-heavy instructional approaches can exacerbate cognitive overload, particularly in conceptually dense domains such as pharmacology [6,12,13]. Learners frequently report difficulty retaining pharmacologic content beyond short-term assessments, with perceived gaps between didactic instruction and applied clinical reasoning [14,15]. These challenges underscore the need for instructional strategies that explicitly support learning and memory rather than mere information exposure.
One promising approach is the use of narrative-based multimedia instruction, often described under the broader umbrella of edutainment [16,17]. Grounded in the Cognitive Theory of Multimedia Learning (CTML), this paradigm posits that learning is enhanced when information is presented through coordinated visual and auditory channels within the limits of human cognitive capacity [12,18]. Well-designed multimedia narratives can promote generative processing by integrating imagery, spoken language, and temporal structure in ways that reduce extraneous cognitive load while supporting schema construction. Although edutainment strategies have been widely deployed in public health communication, such as successful interventions targeting HIV prevention and COVID-19 awareness through music and narrative media, their theoretical underpinnings are directly applicable to medical education contexts [19,20].
Although multimedia resources and case-based learning are now widely incorporated into undergraduate medical curricula, substantial challenges in pharmacology learning persist [13]. Surveys of undergraduate medical education consistently identify pharmacology as a high-burden domain associated with poor confidence and frequent knowledge decay, despite increasing use of multimedia supplements [13,21,22]. These persistent challenges underscore the need to evaluate not simply whether multimedia is used, but how it is designed, sequenced, and aligned with learning theory. Accordingly, the present study evaluates a specific animated Cinematic Clinical Narrative (CCN) which represents a structured application of CTML principles within medical education, embedding core biomedical content into professionally designed narrative short films inspired by mainstream popular culture. Prior work has demonstrated that CCNs can enhance learner engagement and support knowledge acquisition in pharmacology, with students reporting higher preference for CCNs compared with traditional text-based cases and demonstrating comparable or improved examination performance [23,24]. Importantly, CCNs are intentionally designed to integrate visual and linguistic mnemonics, narrative causality, and emotional salience, all of which have been shown to support encoding and retrieval processes central to long-term memory formation [25,26].
The recent emergence of generative artificial intelligence (genAI) tools for video, image, and audio production has expanded the feasibility of developing high-quality CCNs at scale. AI-assisted production of educational materials is beginning to permeate all areas of medical education including AI-generated videos [27], diagnostic images/algorithms [28], practice questions [29], and even podcasts [30]. However, the rapid proliferation of AI-generated educational media has raised concerns regarding accuracy, pedagogical rigor, and the dissemination of superficially polished but educationally unsound content [31]. While such “AI-generated slop” has been documented in biomedical education platforms, expert-designed CCNs differ fundamentally in that they are grounded in established learning theory, developed under disciplinary oversight, and subjected to iterative review for clinical accuracy [23]. In this context, genAI functions as a production tool rather than a source of instructional content, supporting efficiency without displacing expert judgment.
Despite growing interest in CCNs and multimedia instruction, empirical validation of how learners attend to and process embedded educational elements remains limited. Biometric methods such as eye tracking offer an opportunity to objectively examine attentional alignment between multimedia design and learner processing, providing exploratory validation of theoretical design principles [32,33].
Accordingly, the present study was designed to empirically evaluate an animated CCN, Twilight: Breaking Clots, as a complementary instructional approach for teaching anticoagulant pharmacology within undergraduate medical education. Using a mixed-methods, quasi-experimental design, the study examined learning outcomes relative to traditional case-based instruction, assessed learner interest using the Situational Interest Survey–Multimedia (SIS-M), and explored visual attention to embedded mnemonic elements through eye-tracking measures. Together, these outcomes aim to contribute to the growing evidence base supporting CCNs as a rigorously designed, theory-informed, and empirically evaluated instructional adjunct, rather than a replacement for established educational practices.

2. Materials and Methods

2.1. Study Design

A multi-site quasi-experimental design was employed across the WWAMI regional medical education system, a distributed undergraduate medical education program spanning Washington, Wyoming, Alaska, Montana, and Idaho. Students across all sites receive a common core curriculum, instructional materials, and examinations delivered on a synchronized schedule. This design allowed for comparison of instructional approaches while maintaining curricular consistency across sites, a strategy commonly used in medical education research when randomization is impractical. The study was conducted during a six-week Cancer, Hormones, and Blood course, which included a dedicated two-week hematology section. Instruction during this block included a one-hour in-person lecture on anticoagulant pharmacology, during which students were exposed to either a traditional case-based discussion or the Twilight: Breaking Clots CCN (Methods S1), depending on site assignment.
Students from Site 6 comprised the experimental group and received the animated CCN intervention in addition to standard instruction. Students from Sites 1–5 served as the comparison group and received traditional instructor-led, text-based clinical cases without exposure to the animated CCN. Since the animated CCN was delivered during a standard class period by the course faculty, this precluded random assignment or removal of students without disrupting instruction. All other instructional materials, curricular content, and assessments were standardized across sites, such that the inclusion of the animated CCN represented the sole instructional material difference between groups. Additional comparators, such as gamified alternatives, were not included in order to avoid introducing multiple concurrent instructional variables and to enable focused evaluation of the animated CCN as implemented in routine educational practice.

2.2. Participants

Participants were first-year medical students (MS1) enrolled in a six-week Cancer, Hormones, and Blood course that included a two-week hematology block. Participation in pretest and posttest assessments was voluntary. To explore attentional engagement with multimedia mnemonics, a separate convenience sample of second-year medical students (MS2) from Site 6 was recruited for biometric data collection, including eye tracking. MS2 students were not part of the achievement analysis.
This design choice was made to avoid contamination of eye-tracking data. The MS1 experimental cohort viewed the CCN during scheduled class time, which precluded concurrent eye tracking. Conducting eye tracking after class would have required repeat viewing, while conducting it prior to the hematology block would have occurred before formal anticoagulation instruction. MS2 participants had previously completed anticoagulation coursework, allowing for naturalistic viewing without disrupting the instructional sequence or introducing novelty effects related to first exposure.
Participants contributing biometric data received a $10 gift card as compensation for their time.

2.3. Intervention: Twilight: Breaking Clots

The animated CCN short film, Twilight: Breaking Clots (Movie S1), was developed using narrative and aesthetic elements inspired by the Twilight franchise and presented in a Claymation style (Figure 1). A script for the animated CCN was generated with OpenAI’s GPT5. Claymation style character designs and locations were generated independently and then combined into single-shot scenes using Google Gemini’s Nano-Banana. Composited scenes were used as starting frames while script descriptions and dialogue were utilized as prompts to generate animations with Google Flow (Veo 3-fast). Prompt design followed established strategies including role prompting [34,35], structured and constraint-based prompting [36], resampling [37], and reference-conditioned affective prompting [38]. Faculty provided domain-specific outlines, learning constraints, and audience context, while generative models were used to propose multiple candidate outputs for expert selection and refinement. Animation prompts provided guidance on character dialogue, motion, and camera movement to support alignment with the intended narrative and instructional cues. ElevenLabs Voice Changer was used to alter Google Flow dialogue output into consistent voices and their SFX platform was used to produce supplemental sound effects. VocalRemover’s Splitter AI was used to remove the Veo 3-fast voices from the original clips to allow for ElevenLabs Voice Changer audio clips to be added to each clip. It was also used to separate the vocals and instrument tracks from Taylor Swift’s song We are Never Getting Back Together. The instrument track was then altered using Udio 1.5. GPT 4o was used to rewrite the lyrics to be anticoagulant focused, and a human singer sung the lyrics, which was later overlayed back onto the instrumental track. Completed scenes and audio were composited using Adobe Premiere Pro.
Anticoagulation pharmacology content was embedded through the use of symbolic imagery, visual mnemonics, and character-driven metaphors (Table 1). The animated CCN incorporated characters such as the Red Blood Cell and Platelet Medimon (Figure 2), which originate from Medimon, an educational, Pokémon-inspired game previously integrated into components of the WWAMI curriculum [39]. Use of these established characters was intended to promote continuity across curricular elements and capitalize on learner familiarity from prior instructional experiences. This allowed for integration of other parts of the curriculum to provide familiarity with parts of the animated CCN. Specific mechanisms of action for warfarin, direct oral anticoagulants (DOACs), and heparin were portrayed through character behaviors and plot events. Additional pharmacologic concepts including reversal agents, drug interactions, and protein C depletion were integrated using purposeful mnemonic devices. The narrative structure was intentionally designed to align visual metaphors with mechanistic principles to promote conceptual understanding and enhance long-term retention.

2.4. Educational Outcome Measures

Learning outcomes were assessed using a standardized pretest and posttest consisting of eight multiple-choice questions aligned with course learning objectives for anticoagulant pharmacology. All MS1 students completed the pretest during the first week of the hematology block. The posttest was administered two weeks after delivery of the anticoagulant pharmacology lecture and associated instructional materials. The two-week delay between instruction and posttesting was intentionally selected to assess retention beyond immediate recall, rather than short-term performance effects associated with recent exposure, which aligns with prior work [40,41]. Delayed assessments are commonly used in educational research to evaluate more durable learning and better reflect long-term knowledge retention than immediate posttests alone.
Baseline academic comparability between instructional groups was evaluated using site-level average performance across five course examinations unrelated to the CCN intervention. This approach is commonly used in quasi-experimental educational studies to assess equivalence of prior academic performance when random assignment is not feasible.

2.5. Engagement Measures

Learner engagement and situational interest were assessed using the Situational Interest Survey–Multimedia (SIS-M), a validated instrument designed to measure triggered interest, maintained interest, affective engagement, and perceived value in multimedia learning environments [42,43]. The SIS-M has been previously applied in medical education research to compare instructional modalities and assess learner response to multimedia design features [44].
MS1 participants at Site 6 who viewed the animated CCN completed the SIS-M. Students completed the survey twice: once in reference to the animated CCN (Twilight: Breaking Clots) and once in reference to non-animated CCN cases delivered elsewhere in the course using text, static images, and audio (Appendix A Table A1). This comparator was selected over traditional text-based cases to provide a closer instructional contrast by holding narrative structure and multimodal elements constant while varying animation and dynamic visual presentation. Responses were recorded on a five-point Likert scale (1 = strongly disagree to 5 = strongly agree), allowing for within-participant comparison of engagement across presentation modalities. In addition, the SIS-M included a forced-choice preference item asking learners to indicate which case style they preferred overall. Students were also provided with an open-ended prompt inviting them to explain the rationale for their preference in their own words, enabling qualitative analysis of decision-making and perceived value.

2.6. Biometric Measures

Eye-tracking data were collected using the Gazepoint GP3 HD eye tracker (Gazepoint Research Inc., Vancouver, BC, Canada), which samples binocular gaze data at 150 Hz. The eye tracker was mounted below a standard computer monitor, and participants were seated at a fixed viewing distance consistent with the manufacturer’s recommendations. A standard calibration procedure was performed for each participant prior to data collection to ensure accurate gaze tracking.
All participants viewed the Twilight: Breaking Clots animated CCN in its entirety without pausing or interruption. The eye-tracking task consisted of passive viewing, with no additional instructions beyond attending the film as they normally would during an educational video. The participants were educated on the various visual and story mnemonics (Table 1) in advance of viewing the animated CCN to ensure interpretability of gaze patterns rather than to direct attention during viewing.
Eye-tracking data were analyzed using Gazepoint Analysis UX (version 7.1.0). Areas of interest (AOIs) were manually defined around all visual and story-based mnemonic elements embedded within the film (Table 1). AOIs were dynamic, moving and resizing as needed to follow the corresponding mnemonic elements during animation and camera movement. AOIs were present for the full duration that each mnemonic appeared on screen and did not overlap with one another. AOI onset and offset times were recorded to align gaze metrics with the temporal presentation of each mnemonic.
Eye-tracking metrics were calculated at the AOI level for each participant. AOI uptake was defined as the presence of at least one fixation within the AOI. Visual engagement was quantified using total dwell time, defined as the cumulative duration of all fixations within an AOI. To account for differences in on-screen duration across mnemonics, normalized dwell time was calculated by dividing total dwell time by the AOI’s on-screen duration. Discoverability of mnemonics was assessed using time to first fixation (TTFF), calculated relative to AOI onset. Fixation count and revisit count were recorded as secondary indicators of visual processing.
AOI-level statistics were extracted from Gazepoint Analysis UX for each participant. One participant was excluded from analysis due to inadequate eye-tracking data quality. All remaining participants had sufficient tracking quality and viewed the full duration of the film. AOIs associated with end credits were excluded from instructional analyses to focus on educational content.
Eye-tracking outcomes were summarized using descriptive statistics. For each AOI, hit rate was calculated as the proportion of participants who fixated the AOI at least once. Dwell time, normalized dwell time, and TTFF were summarized using medians to reduce the influence of outliers. Analyses were conducted at both the individual AOI level and at the mnemonic concept level, where repeated AOIs representing the same mnemonic across multiple scenes were collapsed to assess cumulative attentional engagement. All analyses were conducted using exported AOI-level data from Gazepoint Analysis UX.

2.7. Data Collection Procedures

The pretest was administered to MS1 students during the week preceding the anticoagulant pharmacology lecture, which included the CCN intervention. Two weeks following the lecture, MS1 students completed the posttest and participants at Site 6 also completed the SIS-M survey. MS2 participants at Site 6 viewed the CCN with concurrent eye-tracking during a similar time window corresponding to the MS1 posttest and SIS-M administration.

2.8. Data Analysis

Achievement outcomes were analyzed using Microsoft Excel (version 2511). Within-group changes from pretest to posttest were evaluated using paired comparisons. Between-group differences were assessed using independent-samples Student’s t-tests, as appropriate. Fold change and normalized change scores were calculated to descriptively characterize relative learning gains while accounting for baseline performance [45].
Learner engagement data collected via the SIS-M were analyzed using Statisty [46]. Likert-scale responses were summarized using descriptive statistics (mean and standard deviation), and the Wilcoxon signed-rank test was used to evaluate differences in perceived interest. Quantitative SIS-M analyses focused on comparing learner engagement with animated CCN versus non-animated CCN formats.
Open-ended SIS-M responses were analyzed using a qualitative thematic analysis approach. Qualitative coding was conducted using the Agentic Thematic Analysis workflow implemented in the Google Opal App, an AI-assisted qualitative analysis pipeline previously developed and reported by our laboratory [28]. The human-in-the-loop workflow enables iterative code generation, refinement, and theme consolidation while maintaining analyst oversight. All AI-generated codes and thematic groupings were reviewed, refined, and finalized by the research team, with disagreements resolved through discussion to ensure interpretive accuracy. This approach supports systematic identification of recurring patterns in learner engagement and enhances transparency and reproducibility in qualitative analysis.

2.9. Ethics

This study (25–127) was approved by the University of Idaho IRB. All SIS-M responses were collected anonymously, and pre/posttest and biometric data were anonymized prior to analysis. Written informed consent was obtained from pre/posttest/SIS-M participants and those contributing biometric data.

3. Results

3.1. Learning Outcomes

Students from six medical school campuses completed a pretest and posttest assessing knowledge of anticoagulant pharmacology. The pretest was administered one week prior to the anticoagulant pharmacology lecture, and the posttest was administered two weeks following the intervention. Students from sites 1–5 did not view the animated CCN and served as the control group (pretest n = 25, posttest n = 22), whereas students from site 6 viewed the animated CCN and served as the experimental group (pretest n = 27, posttest n = 14).
To establish baseline academic comparability between groups, average examination performance across five course exams was examined at the site level. Baseline exam performance was similar across sites, indicating comparable foundational knowledge prior to the intervention (Figure 3a).
Both groups demonstrated significant improvements from pretest to posttest. In the control group, mean scores increased from 44% on the pretest to 73% on the posttest, representing a statistically significant within-group improvement (p < 0.001). Similarly, the experimental group showed a significant increase from a pretest mean of 38% to a posttest mean of 82% (p < 0.001). Despite these within-group gains, there was no statistically significant difference between groups at baseline (pretest p = 0.2580) or on the posttest (p = 0.2535), indicating that absolute scores at each time point were comparable between groups (Figure 3b).
To further examine relative learning gains, fold change (FC) between pretest and posttest scores was calculated for each group. The experimental group demonstrated a substantially larger fold change (313%) compared with the control group (181%), and this difference was statistically significant (p = 0.0143), indicating a greater relative improvement in scores among students who viewed the animated CCN (Figure 3b). This finding suggests that, while absolute posttest scores were similar, the magnitude of improvement from baseline differed between groups.
Normalized change scores (c) were also calculated to account for baseline performance and potential ceiling effects. Analysis of normalized gains revealed no statistically significant difference between the control and experimental groups (Figure 3c,d). Examination of individual performance trajectories showed that only one student in each group scored lower on the posttest compared with the pretest, two students in the control group demonstrated no change in score, and all remaining students in both groups demonstrated improved performance on the posttest. Together, these findings indicate that although normalized gains did not differ between groups, the vast majority of participants exhibited positive learning gains following instruction, with fold-change analysis revealing a greater relative improvement in the experimental group.

3.2. Interest Survey Results

Among the 11 participants that completed the SIS-M, the majority (n = 8, 72.73%) indicated a preference for the Twilight CCN over non-animated cases, while a smaller number (n = 2, 18.18%) preferred the conventional non-animated cases, and one participant expressed no preference (n = 1, 9.09%).
Because participation in the SIS-M survey was limited to students from Site 6 (n = 11), the data were analyzed using descriptive statistics and the Wilcoxon signed-rank test, with attention to four dimensions of situational interest: triggered situational interest (TSI), maintained-feeling interest (MFI), maintained-value interest (MVI), and overall maintained interest (OMI).
Participants reported higher levels of TSI when interacting with the Twilight CCN (M = 4.73, SD = 0.49) compared with non-animation movie cases (M = 4.07, SD = 0.63). A Wilcoxon signed-rank test indicated that participants’ TSI was significantly higher for the Twilight CCN than for the non-animated case (z = 2.40, p = 0.019). The median TSI score on the 5-point Likert scale was 5.00 for the Twilight CCN and 4.25 for the non-animated cases.
Although the OMI was slightly higher for the Twilight CCN (M = 4.57, SD = 0.46) than for the non-animated cases (M = 4.43, SD = 0.59), a Wilcoxon signed-rank test indicated no significant difference between conditions (z = 0.85, p = 0.444). The median OMI score was 4.63 for both the Twilight CCN and the non-animated cases.
With respect to the components of OMI, both maintained-feeling interest (M = 4.39, SD = 0.58) and the maintained-value interest (M = 4.75, SD = 0.42) were higher for the Twilight CCN than for the non-animated cases (M = 4.18, SD = 0.70 and M = 4.68, SD = 0.56, respectively). However, Wilcoxon signed-rank tests indicated that these differences were not statistically significant for either MFI (z = 1.06, p = 0.32) or MVI (z = 0.32, p = 0.833). The median scores on the 5-point Likert scale were 4.50 (MFI) and 5.00 (MVI) for the Twilight CCN and 4.25 (MFI) and 5.00 (MVI) for the non-animated cases.
Thematic analysis of the 11 open-ended survey responses revealed three key themes regarding student perceptions of the animated CCN compared to non-animated instructional formats. These themes elucidate the distinct pedagogical roles students assigned to each format, the affective and motivational impact of animation, and the contextual factors influencing the efficacy of these tools.
  • Theme 1: Distinct Pedagogical Roles: Memorization via Animation vs. Application via Non-Animated Cases.
A predominant theme was the perception that the animated and non-animated formats serve distinct functions within the learning process. Students consistently identified the animated CCN as a superior tool for the initial acquisition and memorization of complex pharmacological information. The narrative structure, combined with integrated visual and auditory mnemonics, was seen as a powerful mechanism for encoding information. As one student explained, “having a moving story in my brain makes it easier for me to recall the information later” (Participant 7). Others highlighted the lasting impact of specific mnemonics, such ‘river rocks the dam’ for rivaroxaban (Participant 1), with one noting, “I still use the film to remember a good chunk of the anticoagulants” (Participant 6).
Conversely, non-animated cases were valued for facilitating higher-order cognitive skills, such as clinical application and reasoning. This format was perceived as a bridge between foundational knowledge and clinical practice. Students described these cases as a “blend [of] case-based learning and application” that better “facilitate[s] deeper learning” (Participant 1). The non-animated format was seen as a practical exercise in clinical reasoning, helpful for “working through [a] step question stem and seeing the symptoms of what the patient would present with and why we would use a certain drug” (Participant 5).
  • Theme 2: Heightened Engagement and Affective Appeal of the Animated Narrative.
The animated CCN elicited a strong positive affective response and was consistently described as more engaging and enjoyable than non-animated cases. Participants frequently used words like “fun” (Participants 1, 8), “interesting,” and “catchy” (Participant 5) to describe their experience. This enhanced engagement was attributed to the novelty of the cinematic format, as one student stated, “Something about it being in film form made it more engaging” (Participant 11). The combination of entertainment and education was seen as a significant benefit for visual learners who found that “the combination of an interesting film with educational aspects really benefited me” (Participant 6). Furthermore, engagement was heightened by the knowledge that the resource was created by peers, with one participant noting, “I think there’s something to knowing that I’m watching the culmination of much effort from fellow med students that demands my attention” (Participant 10). This theme underscores the importance of affective and motivational factors in learning, suggesting that the narrative elements of the animated case effectively capture attention and help the material “stick” (Participant 11).
  • Theme 3: Instructional Design and Complementary Use.
The final theme addresses the practical factors influencing the effectiveness of both formats, leading to the conclusion that they serve complementary roles. Students identified instructional design limitations with the non-animated cases that increased cognitive load, primarily related to rapid pacing that provided “Not enough time to think about question before the [movie] gave you the answer” (Participant 3). Moreover, students felt that deploying application-based cases too soon after a lecture was counterproductive, as they had “not enough time to understand the material enough to know what the movie is talking about” (Participant 3).
In light of the distinct strengths and weaknesses, a synthesizing view emerged wherein students saw the two formats not as mutually exclusive but as “both useful in different ways” (Participant 5). They perceived a staged learning process where the animated CCN is most beneficial for initial learning, while “The non-animated cases are better once you have learned the drugs” (Participant 2). This sentiment was echoed by the student with no preference, who concluded, “both non-animated and animated movie cases give me ways to learn and remember pharmacology content” (Participant 9).

3.3. Biometric Results

Eye-tracking data were collected from 13 s-year medical students (MS2) while they viewed the animated anticoagulant pharmacology CCN. One participant was excluded due to inadequate tracking quality, resulting in a final analytic sample of 12 participants, all of whom viewed the film in its entirety. A total of 45 predefined areas of interest (AOIs), corresponding to visual mnemonics embedded throughout the animated CCN (summarized in Table 1), were analyzed. AOIs dynamically moved and resized to follow mnemonic elements on screen and were present for the full duration of each mnemonic’s appearance.
Overall visual coverage of mnemonic elements was high across viewers. Participants fixated a median of 39.5 out of 45 AOIs (range: 33–42), corresponding to a median coverage rate of 87.8%. This indicates that learners visually sampled the vast majority of instructional mnemonics presented throughout the film.
AOI uptake was operationalized as the proportion of participants who fixated each AOI at least once. After excluding AOIs associated with the end credits, 38 instructional AOIs remained for analysis. Across these AOIs, the median hit rate was 95.8%, with 32 of 38 AOIs viewed by at least 75% of participants and 19 AOIs viewed by all participants. The most consistently viewed AOIs are illustrated in Figure 4, and detailed uptake, dwell, and timing metrics for the top 10 AOIs are summarized in Table 2. These high-uptake AOIs corresponded to visually prominent and narratively emphasized mnemonics embedded within the animated CCN.
Visual salience and processing depth were further examined using dwell time and normalized dwell time. Dwell time was defined as the total duration of fixations within an AOI, while normalized dwell time was calculated by dividing dwell time by AOI on-screen duration to account for differences in exposure length. Across instructional AOIs, the median normalized dwell time was 0.313, indicating that participants spent approximately 31% of each AOI’s on-screen duration visually attending to the mnemonic. Core instructional mnemonics demonstrated both high hit rates and high normalized dwell values, suggesting sustained visual processing rather than brief or incidental glances (Table 2; Figure 4).
Discoverability of visual mnemonics was assessed using time to first fixation (TTFF), calculated relative to AOI onset. Across instructional AOIs, the median TTFF was 0.49 s. Several central mnemonics were fixated within the first 0.1 s of appearing on screen, indicating rapid attentional capture. In contrast, AOIs with longer TTFF values tended to be smaller, more peripheral, or secondary visual elements, suggesting lower attentional priority within complex scenes.
Despite the overall high level of engagement, variability was observed across individual AOIs. A small subset of mnemonics exhibited consistently lower hit rates and reduced dwell times, indicating limited visual engagement. Representative examples of the least-viewed AOIs are shown in Figure 5, with quantitative metrics for the bottom 10 AOIs provided in Table 3. These AOIs were typically peripheral or less narratively emphasized elements within visually dense scenes.
To assess engagement with mnemonic concepts that appeared multiple times throughout the film, AOIs representing the same underlying mnemonic were collapsed and analyzed at the concept level. Core mnemonic concepts demonstrated near-universal uptake and high cumulative dwell times across repeated appearances. The four most salient mnemonic concepts following this aggregation are illustrated in Figure 6, with detailed concept-level metrics presented in Table 4. These results indicate that even when individual AOI instances were occasionally missed, learners reliably engaged with the underlying mnemonic concepts across the film.
Collectively, the eye-tracking findings indicate that visual mnemonics embedded within the animated CCN were consistently noticed, rapidly fixated, and often sustained visual attention across viewers. Most instructional AOIs demonstrated high uptake, with core mnemonics showing particularly strong salience at both the AOI and concept levels. Together, these results provide converging evidence that the visual mnemonic design effectively directed learner attention toward key anticoagulant pharmacology concepts.

4. Discussion

4.1. Summary of Key Findings

This study was intentionally designed as a mixed-methods, quasi-experimental evaluation of an animated Cinematic Clinical Narrative (CCN) as a complementary instructional approach, rather than as a randomized efficacy trial or a test of instructional replacement. Accordingly, the findings should be interpreted as evidence of how CCNs function alongside traditional case-based instruction within authentic curricular settings. By integrating achievement measures, situational interest, learner preference, qualitative explanations, and exploratory biometric data, the present work contributes converging evidence that CCNs can support early knowledge encoding and learner engagement while coexisting with application-focused instructional modalities. This framing aligns with contemporary instructional design principles that emphasize the strategic sequencing and combination of learning tools to support different phases of expertise development, rather than reliance on a single pedagogical format.
This study provides converging quantitative and qualitative evidence that Twilight: Breaking Clots, an animated CCN, functions as a promising and supportive instructional approach for anticoagulant pharmacology within undergraduate medical education. While absolute posttest scores did not differ significantly between groups, fold-change analyses suggested greater relative improvement among students exposed to the CCN, supporting its role as a meaningful adjunct rather than a replacement for traditional instruction. Fold-change analysis was included as a descriptive metric to contextualize relative improvement from baseline, particularly in settings where pre-instruction knowledge differs across cohorts. While normalized change scores are well suited for comparative inference, fold change provides complementary insight into proportional learning gains without implying superiority. Importantly, these achievement findings are complemented by robust engagement data and learner-generated explanations that clarify how and why CCNs may support learning.
Results from the SIS-M demonstrated that the animated CCN elicited significantly higher triggered situational interest compared with non-animated instructional cases. Triggered interest reflects learners’ immediate attentional capture and curiosity in response to instructional stimuli and is considered a critical gateway to deeper engagement and sustained learning [42,47,48]. In contrast, differences between animated and non-animated formats were not statistically significant for maintained interest, affective engagement, or perceived value. This pattern aligns with theoretical models of interest development, which posit that instructional novelty and salience primarily influence early phases of engagement, while deeper interest and value are shaped by repeated exposure and opportunities for application [47,49].
Rather than indicating a limitation of the animated CCN, this profile suggests that CCNs are particularly well suited for initial encoding and motivational priming, while complementary instructional formats may be required to support later-stage consolidation and transfer. Such differentiation is consistent with prior multimedia learning research demonstrating that affectively engaging materials are most effective when integrated into broader instructional sequences rather than used in isolation [18,50].
Thematic analysis of open-ended learner responses provided critical explanatory depth to the quantitative findings. Students consistently articulated a distinction between the pedagogical roles of animated and non-animated instructional formats. The animated CCN was overwhelmingly perceived as superior for memorization and recall, particularly during early learning. Learners described the narrative structure and multimodal mnemonics as creating durable mental representations that facilitated later retrieval, an observation that aligns closely with dual coding theory and narrative transportation frameworks [6,25,26,51]. By embedding pharmacologic mechanisms within a coherent storyline, the CCN appeared to support the construction of integrated schemas rather than isolated factual knowledge.
Conversely, non-animated cases were valued for clinical application and reasoning, particularly once foundational knowledge had been established. Learners framed these cases as opportunities to practice stepwise clinical logic, symptom interpretation, and drug selection, echoing longstanding evidence that case-based learning is most effective when learners possess sufficient prior knowledge to avoid excessive intrinsic cognitive load [1,52]. These findings reinforce the notion that no single instructional format optimally supports all phases of learning, particularly in cognitively demanding domains such as pharmacology.
Importantly, students did not frame the two approaches as competing alternatives. Instead, a synthesized perspective emerged in which animated CCNs were viewed as optimal for early-stage learning and memory formation, while non-animated cases were preferred for subsequent application and refinement. This pattern suggests that the instructional impact of CCNs may be sequencing-dependent, with their greatest value occurring when learners are first encountering complex pharmacologic concepts, rather than after foundational knowledge has already been established. When deployed too late in the learning sequence, narrative and mnemonic supports may offer diminishing returns or be perceived as less relevant to learners focused on clinical reasoning and transfer. This staged learning model mirrors established instructional design principles, including scaffolding and the expertise reversal effect, which emphasize aligning instructional supports with learner readiness [53,54,55]. Accordingly, the findings underscore the importance of considering when CCNs are introduced within a curriculum, rather than evaluating their effectiveness in isolation.
Learners’ descriptions of the animated CCN emphasized heightened enjoyment, novelty, and emotional resonance. Participants frequently characterized the experience as “fun,” “catchy,” and “engaging,” and several noted that the cinematic format increased their willingness to attend closely to the material. Affective engagement has been shown to play a critical role in sustaining effort, reducing perceived cognitive burden, and supporting persistence in challenging learning contexts [56,57,58]. In medical education, where burnout and disengagement are prevalent, instructional approaches that enhance enjoyment without sacrificing rigor are particularly valuable [59,60].
Interestingly, learners also highlighted the social and human dimensions of the CCN, noting that awareness of peer involvement in its creation increased attentional investment. This observation aligns with social learning theory and emerging work on authenticity and relational trust in educational media, suggesting that perceived human effort and intent may amplify engagement even within AI-assisted productions [10,61,62].
Eye-tracking analyses provided exploratory validation that learners visually attended to the intended mnemonic elements embedded within the CCN. High uptake, rapid time to first fixation, and sustained normalized dwell times for core mnemonics suggest that the multimedia design successfully guided learner attention toward high-yield pharmacologic concepts. These findings are consistent with attention-guidance research demonstrating that well-designed visual cues can reduce extraneous cognitive load and promote germane processing [33,63,64].
Importantly, biometric measures were used not to infer learning mechanisms directly, but to assess attentional alignment between instructional intent and learner behavior. This distinction is critical, as eye tracking offers objective evidence of where learners allocate visual attention but cannot, on its own, establish causal pathways to learning outcomes [65,66]. When interpreted alongside achievement and self-report data, however, these measures strengthen the argument that CCNs are not merely entertaining, but pedagogically purposeful.
Within the broader discourse surrounding genAI in education, these findings contribute to an emerging distinction between expert-led, theory-informed AI-assisted instruction and low-quality, unvetted educational media. Unlike so-called “AI-generated slop,” the CCN evaluated in this study was developed through iterative expert review, explicit alignment with learning theory, and empirical evaluation of both outcomes and learner experience. In this context, genAI functioned as a production accelerator rather than a pedagogical decision-maker, preserving instructional integrity while enabling creative expression [23,24,31].
Development of an animated CCN has become increasingly accessible as genAI tools mature. Recent improvements in character consistency allow creators to generate a single character image and then condition subsequent image generation to composite that character into new scenes. These scenes can serve as starting frames for video generation tools, which can additionally generate synchronized dialogue and sound effects to complete individual scenes. In practice, this process may require generating multiple candidate outputs (“rerolling”), although our experience suggests that one to two generations typically yield a usable result. While generating a dedicated soundtrack is not required, background music can enhance narrative immersion. The primary human time investment involves compositing visual, audio, and narrative elements into a cohesive short film, which varies depending on creator experience, film length, and desired complexity. Notably, all AI generation in this study was performed using cloud-based tools without local computational requirements, and final compositing was completed on a standard laptop, supporting the feasibility and accessibility of CCN production for educators with an internet connection and modest technical resources.

4.2. Limitations

This study has several limitations that should be considered when interpreting the findings. The non-randomized, quasi-experimental design limits causal inference and introduces the potential for unmeasured confounding between sites. Because the CCN intervention was implemented at a single site, unmeasured site-specific factors such as instructional culture, faculty delivery style, or learner expectations may have influenced engagement outcomes. Sample sizes, particularly for posttest comparisons and eye-tracking analyses, were modest and may have limited statistical power to detect small between-group differences. Use of the SIS-M was restricted to the experimental site, preventing direct comparison of student self-reported engagement across cohorts. Similarly, biometric data were collected only from Site 6 participants, limiting generalizability of these measures and precluding between-site biometric comparisons. In addition, the study did not include long-term follow-up assessments beyond the two-week delayed posttest, precluding conclusions about extended retention or clinical transfer. Finally, because CCN represents a novel instructional approach for participants, observed differences may partially reflect novelty effects rather than sustained learning or engagement benefits.

4.3. Future Directions

Future research should focus on replicating biometric data collection across multiple sites to enhance generalizability and allow for between-site comparisons of objective engagement measures. Expanding CCN implementation across additional pharmacology domains would help determine whether observed benefits extend beyond a single content area. Longitudinal studies examining long-term knowledge retention and transfer are also needed to assess the durability of learning gains associated with CCNs. Finally, integrating CCNs into blended and self-paced curricular formats may clarify their utility in diverse educational settings and inform scalable implementation strategies.

5. Conclusions

Taken together, these results support animated CCNs as a validated and complementary instructional approach that can enhance engagement, support memory encoding, and integrate effectively within staged learning designs. Future work should examine longitudinal retention, optimize sequencing of animated and non-animated cases, and extend biometric validation across additional institutions and content areas. As AI-enabled media production becomes increasingly accessible, establishing evidence-based models for responsible, theory-driven use will be essential for advancing educational quality rather than novelty.

Supplementary Materials

The following supporting information can be downloaded at: https://doi.org/10.5281/zenodo.18133598, Method S1: Anticoagulant lecture slides; Movie S1: Twilight: Breaking Clots CCN.

Author Contributions

Conceptualization, A.L., K.D. and T.B.; data curation, T.B.; formal analysis, M.G. and T.B.; investigation, A.L., K.D. and T.B.; methodology, T.B.; supervision, T.B.; visualization, T.B.; writing—original draft, A.L., K.D. and T.B.; writing—review and editing, M.G. and T.B. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the University of Idaho WWAMI Medical Education Program and the Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health under Grant #P20GM103408.

Institutional Review Board Statement

This study was approved by the institutional review board of the University of Idaho (25-127), 23 June 2025.

Informed Consent Statement

Informed consent was obtained from all subjects involved in the study.

Data Availability Statement

The datasets presented in this article are not readily available because of the sensitive nature of students’ grades. Requests to access the datasets should be directed to Tyler Bland (tbland@uidaho.edu).

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
genAIGenerative Artificial Intelligence
CCNCinematic Clinical Narrative
SIS-MSituational Interest Survey for Multimedia
TSITriggered Situational Interest
OMIOverall Maintained Interest
MFIMaintained Feeling Interest
MVIMaintained Value Interest
WWAMIWashington, Wyoming, Alaska, Montana, Idaho
LLMLarge Language Model
MCQMultiple-Choice Question

Appendix A

Table A1. SIS items. The “case style” was replaced with “Twilight short film” and “non-animated movie cases” for the Twilight Breaking Clots and text/image/audio-based movie cases survey, respectively.
Table A1. SIS items. The “case style” was replaced with “Twilight short film” and “non-animated movie cases” for the Twilight Breaking Clots and text/image/audio-based movie cases survey, respectively.
SIS TypeSurvey Item
SI-triggeredThe “case style” was interesting.
The “case style” grabbed my attention.
The “case style” was often entertaining.
The “case style” was so exciting, it was easy to pay attention.
SI-maintained-feelingWhat I learned from the “case style” is fascinating to me.
I am excited about what I learned from the “case style”.
I like what I learned from the “case style”.
I found the information from the “case style” interesting.
SI-maintained-valueWhat I studied in the “case style” is useful for me to know.
The things I studied in the “case style” are important to me.
What I learned from the “case style” can be applied to my major/career.
I learned valuable things from the “case style”.

References

  1. Van Merriënboer, J.J.G.; Sweller, J. Cognitive load theory in health professional education: Design principles and strategies. Med. Educ. 2010, 44, 85–93. [Google Scholar] [CrossRef]
  2. Young, J.Q.; Van Merrienboer, J.; Durning, S.; Cate, O.T. Cognitive Load Theory: Implications for medical education: AMEE Guide No. 86. Med. Teach. 2014, 36, 371–384. [Google Scholar] [CrossRef] [PubMed]
  3. Hirsh, J.; Dalen, J.E.; Anderson, D.R.; Poller, L.; Bussey, H.; Ansell, J.; Deykin, D. Oral Anticoagulants: Mechanism of Action, Clinical Effectiveness, and Optimal Therapeutic Range. Chest 2001, 119, 8S–21S. [Google Scholar] [CrossRef] [PubMed]
  4. Heidbuchel, H.; Verhamme, P.; Alings, M.; Antz, M.; Diener, H.-C.; Hacke, W.; Oldgren, J.; Sinnaeve, P.; Camm, A.J.; Kirchhof, P.; et al. Updated European Heart Rhythm Association Practical Guide on the use of non-vitamin K antagonist anticoagulants in patients with non-valvular atrial fibrillation. Ep Eur. 2015, 17, 1467–1507. [Google Scholar] [CrossRef] [PubMed]
  5. Sweller, J. Cognitive Load Theory. Psychol. Learn. Motiv.—Adv. Res. Theory 2011, 55, 37–76. [Google Scholar] [CrossRef]
  6. Dunlosky, J.; Rawson, K.A.; Marsh, E.J.; Nathan, M.J.; Willingham, D.T. Improving students’ learning with effective learning techniques: Promising directions from cognitive and educational psychology. Psychol. Sci. Public Interest 2013, 14, 4–58. [Google Scholar] [CrossRef]
  7. Budnitz, D.S.; Pollock, D.A.; Weidenbach, K.N.; Mendelsohn, A.B.; Schroeder, T.J.; Annest, J.L. National Surveillance of Emergency Department Visits for Outpatient Adverse Drug Events. JAMA 2006, 296, 1858–1866. [Google Scholar] [CrossRef]
  8. Piazza, G.; Nguyen, T.N.; Cios, D.; Labreche, M.; Hohlfelder, B.; Fanikos, J.; Fiumara, K.; Goldhaber, S.Z. Anticoagulation-associated adverse drug events. Am. J. Med. 2011, 124, 1136–1142. [Google Scholar] [CrossRef]
  9. Boshuizen, H.P.A.; Schmidt, H. The development of clinical reasoning expertise. In Clinical Reasoning in the Health Professions; Elsevier: Amsterdam, The Netherlands, 2024; pp. 57–65. [Google Scholar]
  10. Bandura, A.; Walters, R. Social Learning Theory; Prentice-Hall: Upper Saddle River, NJ, USA, 1976. [Google Scholar]
  11. Monrouxe, L.V.; Bullock, A.; Gormley, G.; Kaufhold, K.; Kelly, N.; Roberts, C.E.; Mattick, K.; Rees, C. New graduate doctors’ preparedness for practice: A multistakeholder, multicentre narrative study. BMJ Open 2018, 8, e023146. [Google Scholar] [CrossRef]
  12. Mayer, R.E. Cognitive Theory of Multimedia Learning. In The Cambridge Handbook of Multimedia Learning; Cambridge University Press: Cambridge, UK, 2012; pp. 31–48. [Google Scholar]
  13. Fasinu, P.S.; Wilborn, T.W. Pharmacology education in the medical curriculum: Challenges and opportunities for improvement. Pharmacol. Res. Perspect. 2024, 12, e1178. [Google Scholar] [CrossRef]
  14. Brinkman, D.J.; Tichelaar, J.; Graaf, S.; Otten, R.H.J.; Richir, M.C.; van Agtmael, M.A. Do final-year medical students have sufficient prescribing competencies? A systematic literature review. Br. J. Clin. Pharmacol. 2018, 84, 615–635. [Google Scholar] [CrossRef] [PubMed]
  15. Tobaiqy, M. Exploring medical students’ preferences and challenges in clinical pharmacology education: Insights and improvement strategies. BMC Med. Educ. 2025, 25, 374. [Google Scholar] [CrossRef] [PubMed]
  16. Aksakal, N. Theoretical View to The Approach of The Edutainment. Procedia Soc. Behav. Sci. 2015, 186, 1232–1239. [Google Scholar] [CrossRef]
  17. Sokolova, A.; Deviatnikova, K. Edutainment as a New Educational Technology: A Comparative Analysis. Lect. Notes Netw. Syst. 2023, 830, 302–318. [Google Scholar] [CrossRef]
  18. Mayer, R.E. Introduction to Multimedia Learning. In The Cambridge Handbook of Multimedia Learning; Cambridge University Press: Cambridge, UK, 2005; pp. 1–16. [Google Scholar]
  19. Singhal, A.; Rogers, E.M. Entertainment-Education: A Communication Strategy for Social Change; Taylor and Francis: Abingdon, UK, 2012; Available online: https://www.taylorfrancis.com/books/mono/10.4324/9781410607119/entertainment-education-arvind-singhal-everett-rogers (accessed on 28 December 2025).
  20. Dzinamarira, T.; Nachipo, B.; Nyathi, A.; Madziva, R.; Herrera, H.; Siegel, H.; Musuka, G. The case to scale up edutainment as an effective public health communication intervention to combat the COVID-19 pandemic in Zimbabwe. Health Promot. Perspect. 2022, 12, 34–36. [Google Scholar] [CrossRef]
  21. Tekeş, E.; Güngör, B.; Sılan, C.; Toraman, Ç. Enhancing pharmacology education through role-play: Impact on student attitudes. BMC Med. Educ. 2025, 25, 1306. [Google Scholar] [CrossRef]
  22. Fischbein, R.; Bonfine, N. Pharmacy and Medical Students’ Mental Health Symptoms, Experiences, Attitudes and Help-Seeking Behaviors. Am. J. Pharm. Educ. 2019, 83, 7558. [Google Scholar] [CrossRef]
  23. Worthley, B.; Guo, M.; Sheneman, L.; Bland, T. Antiparasitic Pharmacology Goes to the Movies: Leveraging Generative AI to Create Educational Short Films. AI 2025, 6, 60. [Google Scholar] [CrossRef]
  24. Bland, T. Enhancing Medical Student Engagement Through Cinematic Clinical Narratives: Multimodal Generative AI-Based Mixed Methods Study. JMIR Med. Educ. 2025, 11, e63865. [Google Scholar] [CrossRef]
  25. Clark, J.M.; Paivio, A. Dual coding theory and education. Educ. Psychol. Rev. 1991, 3, 149–210. [Google Scholar] [CrossRef]
  26. Paivio, A. Dual coding theory: Retrospect and current status. Can. J. Psychol./Rev. Can. Psychol. 1991, 45, 255–287. [Google Scholar] [CrossRef]
  27. Artsi, Y.; Sorin, V.; Glicksberg, B.; Korfiatis, P.; Thomas, D.C.; Nadkarni, G.N.; Klang, E. AI-generated videos in medical education: Systematic review. BMJ Open Qual. 2025, 14, e003704. [Google Scholar] [CrossRef] [PubMed]
  28. Massaad, C.; Howe, H.; Guo, M.; Bland, T. A Multi-Institution Mixed Methods Analysis of a Novel Acid-Base Mnemonic Algorithm. Multimodal Technol. Interact. 2025, 9, 113. [Google Scholar] [CrossRef]
  29. Law, A.K.; So, J.; Lui, C.T.; Choi, Y.F.; Cheung, K.H.; Hung, K.K.-C.; Graham, C.A. AI versus human-generated multiple-choice questions for medical education: A cohort study in a high-stakes examination. BMC Med. Educ. 2025, 25, 208. [Google Scholar] [CrossRef] [PubMed]
  30. Paudel, K.R. AI-Generated Podcasts as a Supplemental Pharmacology Learning Tool: A Feasibility Study in Undergraduate Medical Education. Med. Sci. Educ. 2025, 2025, 1–7. [Google Scholar] [CrossRef]
  31. Jones, E.M.; Newman, J.D.; Kim, B.; Fogle, E.J. AI-Generated “Slop” in Online Biomedical Science Educational Videos: Mixed Methods Study of Prevalence, Characteristics, and Hazards to Learners and Teachers. JMIR Med. Educ. 2025, 11, e80084. [Google Scholar] [CrossRef]
  32. Hernandez-de-Menendez, M.; Morales-Menendez, R.; Escobar, C.A.; Arinez, J. Biometric applications in education. Int. J. Interact. Des. Manuf. 2021, 15, 365. [Google Scholar] [CrossRef]
  33. Van Gog, T.; Scheiter, K. Eye tracking as a tool to study and enhance multimedia learning. Learn. Instr. 2010, 20, 95–99. [Google Scholar] [CrossRef]
  34. Van Buren, D. Guided scenarios with simulated expert personae: A remarkable strategy to perform cognitive work. arXiv 2023, arXiv:2306.03104. [Google Scholar] [CrossRef]
  35. Wang, R.; Mi, F.; Chen, Y.; Xue, B.; Wang, H.; Zhu, Q.; Wong, K.-F.; Xu, R. Role Prompting Guided Domain Adaptation with General Capability Preserve for Large Language Models. In Proceedings of the Findings of the Association for Computational Linguistics: NAACL 2024—Findings 2024, Mexico City, Mexico, 16–21 June 2024; pp. 2243–2255. [Google Scholar]
  36. Rath, A. Structured Prompting and Feedback-Guided Reasoning with LLMs for Data Interpretation. arXiv 2025, arXiv:2505.01636. [Google Scholar] [CrossRef]
  37. Chen, B.; Zhang, Z.; Langrené, N.; Zhu, S. Unleashing the potential of prompt engineering for large language models. Patterns 2025, 6, 101260. [Google Scholar] [CrossRef] [PubMed]
  38. Li, C.; Wang, J.; Zhang, Y.; Zhu, K.; Hou, W.; Lian, J.; Luo, F.; Yang, Q.; Xie, X. Large Language Models Understand and Can be Enhanced by Emotional Stimuli. arXiv 2023, arXiv:2307.11760. [Google Scholar] [CrossRef]
  39. Bland, T.; Guo, M. Visual Mnemonics and Gamification: A New Approach to Teaching Muscle Physiology. J. Technol.-Integr. Lessons Teach. 2024, 3, 73–82. [Google Scholar] [CrossRef]
  40. Duchastel, P.C. Retention of prose following testing with different types of tests. Contemp. Educ. Psychol. 1981, 6, 217–226. [Google Scholar] [CrossRef]
  41. Karpicke, J.D.; Roediger, H.L. The critical importance of retrieval for learning. Science 2008, 319, 966–968. [Google Scholar] [CrossRef]
  42. Dousay, T.A. Effects of redundancy and modality on the situational interest of adult learners in multimedia learning. Educ. Technol. Res. Dev. 2016, 64, 1251–1271. [Google Scholar] [CrossRef]
  43. Dousay, T.A.; Trujillo, N.P. An examination of gender and situational interest in multimedia learning environments. Br. J. Educ. Technol. 2019, 50, 876–887. [Google Scholar] [CrossRef]
  44. Bland, T.; Guo, M.; Dousay, T.A. Multimedia design for learner interest and achievement: A visual guide to pharmacology. BMC Med. Educ. 2024, 24, 113. [Google Scholar] [CrossRef]
  45. Marx, J.D.; Cummings, K. Normalized change. Am. J. Phys. 2007, 75, 87–91. [Google Scholar] [CrossRef]
  46. Free Statistics App: T-Test, Chi-Square, Correlation, ANOVA, Regression. Available online: https://statisty.app/ (accessed on 1 January 2026).
  47. Renninger, K.A.; Hidi, S. Revisiting the conceptualization, measurement, and generation of interest. Educ. Psychol. 2011, 46, 168–184. [Google Scholar] [CrossRef]
  48. Hidi, S.; Ann Renninger, K. The four-phase model of interest development. Educ. Psychol. 2006, 41, 111–127. [Google Scholar] [CrossRef]
  49. Guo, J.; An, F. Exploring the categories of students’ interest and their relationships with deep learning in technology supported environments. Sci. Rep. 2025, 15, 10370. [Google Scholar] [CrossRef] [PubMed]
  50. Moreno, R.; Mayer, R. Interactive Multimodal Learning Environments. Educ. Psychol. Rev. 2007, 19, 309–326. [Google Scholar] [CrossRef]
  51. Green, M.C.; Brock, T.C. The role of transportation in the persuasiveness of public narratives. J. Pers. Soc. Psychol. 2000, 79, 701–721. [Google Scholar] [CrossRef]
  52. Schmidt, H.G.; Rotgans, J.I.; Yew, E.H.J. The process of problem-based learning: What works and why. Med. Educ. 2011, 45, 792–806. [Google Scholar] [CrossRef]
  53. Bransford, J.D.; Brown, A.L.; Cocking, R.R. How People Learn: Brain, Mind, Experience, and School: Expanded Edition; National Academy Press: Washington, DC, USA, 2000. [Google Scholar] [CrossRef]
  54. van Merriënboer, J.J.G.; Kirschner, P.A. Ten Steps to Complex Learning: A Systematic Approach to Four-Component Instructional Design; Routledge: London, UK, 2017. [Google Scholar] [CrossRef]
  55. Kalyuga, S.; Ayres, P.; Chandler, P.; Sweller, J. The expertise reversal effect. Educ. Psychol. 2003, 38, 23–31. [Google Scholar] [CrossRef]
  56. Pekrun, R.; Goetz, T.; Titz, W.; Perry, R.P. Academic emotions in students’ self-regulated learning and achievement: A program of qualitative and quantitative research. Educ. Psychol. 2002, 37, 91–105. [Google Scholar] [CrossRef]
  57. Um, E.; Plass, J.L.; Hayward, E.O.; Homer, B.D. Emotional design in multimedia learning. J. Educ. Psychol. 2012, 104, 485–498. [Google Scholar] [CrossRef]
  58. Ryan, R.M.; Deci, E.L. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions. Contemp. Educ. Psychol. 2000, 25, 54–67. [Google Scholar] [CrossRef]
  59. Cook, D.A.; Artino, A.R. Motivation to learn: An overview of contemporary theories. Med. Educ. 2016, 50, 997–1014. [Google Scholar] [CrossRef]
  60. Dyrbye, L.; Shanafelt, T. A narrative review on burnout experienced by medical students and residents. Med. Educ. 2016, 50, 132–149. [Google Scholar] [CrossRef]
  61. Norman, G. Research in medical education: Three decades of progress. BMJ Br. Med. J. 2002, 324, 1560. [Google Scholar] [CrossRef]
  62. Daher, R. Integrating AI literacy into teacher education: A critical perspective paper. Discov. Artif. Intell. 2025, 5, 217. [Google Scholar] [CrossRef]
  63. De Koning, B.B.; Tabbers, H.K.; Rikers, R.M.J.P.; Paas, F. Towards a Framework for Attention Cueing in Instructional Animations: Guidelines for Research and Design. Educ. Psychol. Rev. 2009, 21, 113–140. [Google Scholar] [CrossRef]
  64. Mayer, R.E.; Fiorella, L. Principles for Reducing Extraneous Processing in Multimedia Learning: Coherence, Signaling, Redundancy, Spatial Contiguity, and Temporal Contiguity Principles. In The Cambridge Handbook of Multimedia Learning, 2nd ed.; Cambridge University Press: Oxford, UK, 2014; pp. 279–315. [Google Scholar]
  65. Jarodzka, H.; Skuballa, I.; Gruber, H. Eye-Tracking in Educational Practice: Investigating Visual Perception Underlying Teaching and Learning in the Classroom. Educ. Psychol. Rev. 2020, 33, 1–10. [Google Scholar] [CrossRef]
  66. Holmqvist, K.; Nyström, M.; Andersson, R.; Dewhurst, R.; Jarodzka, H.; Van de Weijer, J. Eye Tracking: A Comprehensive Guide to Methods and Measures; Oxford University Press: Oxford, UK, 2011. [Google Scholar]
Ai 07 00059 g001
Figure 1. Movie poster for the CCN.
Figure 1. Movie poster for the CCN.
Ai 07 00059 g001
Ai 07 00059 g002
Figure 2. Medimon characters included in the CCN (not including the Clot Medimon in the post credit scene). (a) Red Blood Cell Medimon. (b) Platelet Medimon. Visual mnemonics of each character are labeled.
Figure 2. Medimon characters included in the CCN (not including the Clot Medimon in the post credit scene). (a) Red Blood Cell Medimon. (b) Platelet Medimon. Visual mnemonics of each character are labeled.
Ai 07 00059 g002
Ai 07 00059 g003
Figure 3. Achievement results. (a) Average exam scores across all 6 sites. Site 6 is the experimental site that viewed the Twilight CCN. Data is represented as the mean ± SD. (b) Pre/posttest scores for both the control and CCN groups. (c,d) Normalized change scores for both groups; positive values denote increases from pretest to posttest, whereas negative values denote decreases. Data is represented as the mean, with error bars representing the min/max. *** p < 0.001. c, Normalized change score.
Figure 3. Achievement results. (a) Average exam scores across all 6 sites. Site 6 is the experimental site that viewed the Twilight CCN. Data is represented as the mean ± SD. (b) Pre/posttest scores for both the control and CCN groups. (c,d) Normalized change scores for both groups; positive values denote increases from pretest to posttest, whereas negative values denote decreases. Data is represented as the mean, with error bars representing the min/max. *** p < 0.001. c, Normalized change score.
Ai 07 00059 g003
Ai 07 00059 g004
Figure 4. Representative examples of highly viewed visual mnemonic areas of interest (AOIs) with aggregated eye-tracking data overlaid. Panels show (a) the Heparin scoreboard, (b) necrotic burns, (c) Dexanet, and (d) the War CrYmes (warrant) mnemonic scenes. Colored fixation markers represent individual gaze data for each participant, with marker size proportional to fixation duration, illustrating relative visual attention density within each AOI. Change in gaze location is shown with connecting lines. Software text overlays display the raw data, which is better viewed in the summary shown in Table 2. These visualizations demonstrate consistent attentional allocation to intended mnemonic elements embedded within the animated CCN and provide qualitative evidence of alignment between instructional design and learner visual attention.
Figure 4. Representative examples of highly viewed visual mnemonic areas of interest (AOIs) with aggregated eye-tracking data overlaid. Panels show (a) the Heparin scoreboard, (b) necrotic burns, (c) Dexanet, and (d) the War CrYmes (warrant) mnemonic scenes. Colored fixation markers represent individual gaze data for each participant, with marker size proportional to fixation duration, illustrating relative visual attention density within each AOI. Change in gaze location is shown with connecting lines. Software text overlays display the raw data, which is better viewed in the summary shown in Table 2. These visualizations demonstrate consistent attentional allocation to intended mnemonic elements embedded within the animated CCN and provide qualitative evidence of alignment between instructional design and learner visual attention.
Ai 07 00059 g004
Ai 07 00059 g005
Figure 5. Representative examples of the least viewed visual mnemonic AOIs. (a) CYP Drone (left), (b,c) X Truck (2) and (3), and (d) FVII. Colored fixation markers represent individual gaze data for each participant, with marker size proportional to fixation duration, illustrating relative visual attention density within each AOI. Change in gaze location is shown with connecting lines. Software text overlays display the raw data, which is better viewed in summary shown in Table 3. These visualizations demonstrate the lack of consistent attentional allocation to intended mnemonic elements embedded within the animated CCN and provide qualitative evidence of misalignment between instructional design and learner visual attention.
Figure 5. Representative examples of the least viewed visual mnemonic AOIs. (a) CYP Drone (left), (b,c) X Truck (2) and (3), and (d) FVII. Colored fixation markers represent individual gaze data for each participant, with marker size proportional to fixation duration, illustrating relative visual attention density within each AOI. Change in gaze location is shown with connecting lines. Software text overlays display the raw data, which is better viewed in summary shown in Table 3. These visualizations demonstrate the lack of consistent attentional allocation to intended mnemonic elements embedded within the animated CCN and provide qualitative evidence of misalignment between instructional design and learner visual attention.
Ai 07 00059 g005
Ai 07 00059 g006
Figure 6. Representative examples of the most salient mnemonic concepts after collapsing repeated AOIs across scenes. (a) Dexanet, (b) Necrotic Burn, (c) Hairpins, and (d) K-Bomb.
Figure 6. Representative examples of the most salient mnemonic concepts after collapsing repeated AOIs across scenes. (a) Dexanet, (b) Necrotic Burn, (c) Hairpins, and (d) K-Bomb.
Ai 07 00059 g006
Table 1. Visual and story mnemonics and their meanings. Colored text in the Mnemonic column correspond to the same-colored text in the Meaning column.
Table 1. Visual and story mnemonics and their meanings. Colored text in the Mnemonic column correspond to the same-colored text in the Meaning column.
Visual/Story MnemonicMeaning
The werewolves have dammed the river which is causing the town and crops to die.A clot obstructing an artery can cause ischemia and tissue death.
Bella attacks the dam with hairpins, which the werewolves block with Proh Solvite shieldsHeparin is an anticoagulant that can be reversed with protamine sulfate.
Edward HITs fastballs at the dam, which only cause the Platelet Medimon to fall off.An adverse effect of heparin treatment is heparin-induced thrombocytopenia (HIT) causing a decrease in platelet levels.
Edward calls in an “X” truck to let the river rock the dam, but is blocked by a Dexa Net from Jacob the alpha wolf.Rivaroxaban is a direct factor Xa inhibitor which can be reversed with Andexanet alpha.
Edward and Bella declare war on the werewolves and use a K-bomb received from the Volturi (numbered robes)Warfarin inhibits vitamin K metabolism, leading to decrease in factors II, VII, IX, and X.
The K-bomb causes the town to be flooded and Bella to get necrotic burns.Warfarin adverse effects included increased risk of bleeding and warfarin-induced skin necrosis.
Edward and Bella are arrested by CYP robots and charged with war CrYmes for flooding the town.CYP2C9 mutations can cause increased levels of warfarin leading to increased risk of bleeding.
Table 2. Top 10 visual mnemonic AOIs ranked by attentional uptake and salience. AOIs are ordered by hit rate (the proportion of participants with at least one fixation within the AOI), followed by normalized dwell time (median dwell time divided by AOI duration), reflecting sustained attention relative to on-screen exposure. Additional metrics include time to first fixation (TTFF), representing how quickly an AOI attracted visual attention after appearing; median dwell time per AOI duration, indicating relative attentional engagement; and median fixation count, reflecting repeated visual sampling of the mnemonic. Together, these metrics characterize discoverability, consistency of attention, and depth of visual engagement with instructional mnemonics.
Table 2. Top 10 visual mnemonic AOIs ranked by attentional uptake and salience. AOIs are ordered by hit rate (the proportion of participants with at least one fixation within the AOI), followed by normalized dwell time (median dwell time divided by AOI duration), reflecting sustained attention relative to on-screen exposure. Additional metrics include time to first fixation (TTFF), representing how quickly an AOI attracted visual attention after appearing; median dwell time per AOI duration, indicating relative attentional engagement; and median fixation count, reflecting repeated visual sampling of the mnemonic. Together, these metrics characterize discoverability, consistency of attention, and depth of visual engagement with instructional mnemonics.
AOIHit Rate
(%)		Viewed (n)Median TTFF
(s, rel)		Median Dwell/
AOI Dur		Median Dwell (s)Median Fixations
K-Bomb Scoreboard100.012.00.2500.7432.2408.0
Necrotic Burn (2)100.012.00.3810.5333.0348.0
Dexanet (2)100.012.00.0000.5323.77811.0
War CrYmes (warrant)100.012.00.0730.4752.0944.5
Dexanet (1)100.012.00.0620.4343.1629.0
K-Bomb (3)100.012.00.9840.4082.2484.0
Heparin Score Board100.012.00.0150.3973.10610.0
HIT (1)100.012.00.5620.3901.2544.0
Platelet Strike (2)100.012.00.4450.3871.3433.0
K-Bomb (2)100.012.00.5320.3420.8082.5
Table 3. Bottom 10 visual mnemonic AOIs ranked by attentional uptake and salience. AOIs are ordered by hit rate (the proportion of participants with at least one fixation within the AOI), followed by normalized dwell time (median dwell time divided by AOI duration), reflecting lack of sustained attention relative to on-screen exposure. Additional metrics include time to first fixation (TTFF), representing how quickly an AOI attracted visual attention after appearing; median dwell time per AOI duration, indicating relative attentional engagement; and median fixation count, reflecting repeated visual sampling of the mnemonic. Together, these metrics characterize discoverability, consistency of attention, and depth of visual engagement with instructional mnemonics.
Table 3. Bottom 10 visual mnemonic AOIs ranked by attentional uptake and salience. AOIs are ordered by hit rate (the proportion of participants with at least one fixation within the AOI), followed by normalized dwell time (median dwell time divided by AOI duration), reflecting lack of sustained attention relative to on-screen exposure. Additional metrics include time to first fixation (TTFF), representing how quickly an AOI attracted visual attention after appearing; median dwell time per AOI duration, indicating relative attentional engagement; and median fixation count, reflecting repeated visual sampling of the mnemonic. Together, these metrics characterize discoverability, consistency of attention, and depth of visual engagement with instructional mnemonics.
AOIHit Rate (%)Viewed (n)Median TTFF
(s, rel)		Median Dwell/
AOI Dur		Median Dwell (s)Median Fixations
CYP Drone (left)8.31.00.6130.3010.4521.0
X Truck (2)33.34.00.4790.1570.3641.0
X Truck (3)50.06.00.8580.0880.1701.0
FVII (1)50.06.00.5440.3130.8652.0
FVII (2)58.37.00.7090.3031.6164.0
FIX (2)66.78.00.6610.1420.6242.0
Proh Solvite Shield (3)75.09.02.9990.0520.2751.0
Hairpin (right)75.09.01.1070.1230.5022.0
Proh Solvite Shield (1)75.09.00.6870.2121.1324.0
FX (2)83.310.00.6980.0460.3642.0
Table 4. Mnemonic concept analysis (collapsing repeated AOIs across scenes).
Table 4. Mnemonic concept analysis (collapsing repeated AOIs across scenes).
Mnemonic ConceptHit Rate (%)Viewed (n)Median Total Dwell (s)Median First TTFF (s, rel)
Dexanet100.0127.4620.000
Necrotic Burn100.0126.4500.381
Hairpins91.7114.4870.084
K-Bomb100.0124.3510.229
Heparin Score Board100.0123.1060.015
Proh Solvite Shield100.0123.0890.096
Platelet Medimon83.3102.8220.013
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Lee, A.; DeWitt, K.; Guo, M.; Bland, T. A Multisite Study of an Animated Cinematic Clinical Narrative for Anticoagulant Pharmacology Education. AI 2026, 7, 59. https://doi.org/10.3390/ai7020059

AMA Style

Lee A, DeWitt K, Guo M, Bland T. A Multisite Study of an Animated Cinematic Clinical Narrative for Anticoagulant Pharmacology Education. AI. 2026; 7(2):59. https://doi.org/10.3390/ai7020059

Chicago/Turabian Style

Lee, Amanda, Kyle DeWitt, Meize Guo, and Tyler Bland. 2026. "A Multisite Study of an Animated Cinematic Clinical Narrative for Anticoagulant Pharmacology Education" AI 7, no. 2: 59. https://doi.org/10.3390/ai7020059

APA Style

Lee, A., DeWitt, K., Guo, M., & Bland, T. (2026). A Multisite Study of an Animated Cinematic Clinical Narrative for Anticoagulant Pharmacology Education. AI, 7(2), 59. https://doi.org/10.3390/ai7020059

Article Metrics

Back to TopTop