Educational Methodology

The problem NEXO solves

Anyone in training across the health professions faces three pressures at once: high curricular volume, fragmented time, and constant exposure to interfaces that compete for attention. The literature calls the result cognitive dead time: windows too short to open a dense chapter, too long to be irrelevant. Microlearning, defined by Buchem and Hamelmann (2010)^[1] as learning units of a few seconds to 15 minutes, was conceived precisely for that window.

The scoping review by De Gagne et al.^[2], published in 2019 in JMIR Medical Education, analyzed 17 studies of microlearning in health professions education and concluded that the strategy consistently improves knowledge and confidence in asynchronous, self-directed sessions. NEXO operates inside that frame. Each case completes in 5 to 10 minutes, is asynchronous, is independent of the others, and the clue hierarchy works as structured retrieval practice.

Three theoretical pillars

Microlearning

Buchem and Hamelmann^[1] describe ten concepts that characterize a microlearning intervention: learning context, time spent, content type, content creation, aggregation, retrieval, structure of the learning cycle, target group, learner role, and participation. In an internal classification we are preparing for publication, NEXO satisfies 7 of the 10 concepts in full, 1 partially, and 2 are deliberately absent for medical safety reasons.

Gamification

The systematic review by van Gaalen et al. (2021)^[3] evaluated 44 studies of gamification in health professions education. The central finding was that the combination of assessment and challenge attributes from Landers' framework^[4] is the most common, present in 24 of the 44 studies, and is associated with increased use of learning material and, in some studies, improved learning outcomes. That combination is the backbone of NEXO: XP per correct answer, country and global leaderboards, sinapses cost on errors, and continent unlocking through cumulative performance.

Game-based learning

The narrative review by Xu et al. (2023)^[5] distinguishes serious games from gamification. NEXO fits as gamification applied to a diagnostic reasoning app, not as a traditional serious game. The study reports that game-based learning improves engagement, long-term retention, and clinical decision-making, three direct targets of our architecture.

NEXO mechanics translated into theory

How each mechanic works in detail

Retrieval practice across six clues

The most durable way to learn is to attempt to retrieve information before reviewing it. Karpicke and Roediger (2008)^[6] showed that test repetition produces long-term retention superior to rereading, and the systematic review by Trumble et al. (2024)^[7] in health professions education synthesizes studies on distributed practice and retrieval practice. The six-clue structure of NEXO is a cascade of retrieval practice. Clue 1 gives only age, sex, and chief complaint: the player is forced to generate differentials with intentionally incomplete information. With each new clue, the player must update the hypothesis, discard differentials, and converge. That exact behavior is what the diagnostic reasoning literature calls iterative hypothesis refinement.

Active retrieval, not passive memorization

Flashcards train factual recall efficiently: you see side A, you retrieve side B. The problem is that medical exams rarely have that structure. The review by Serra et al. (2025)^[21] on retrieval practice in health professions education shows that isolated flashcards train decontextualized retrieval and stumble when the learner needs to reason through an applied case: relating findings, comparing differentials, choosing management. The clinical case demands contextual retrieval, not single-key retrieval. NEXO was designed to fill that gap. Each clue forces contextual retrieval, the outcome demands a defensible diagnostic choice, and the error enters the flow as a pedagogical signal, never as punishment. Fischer et al. (2006)^[22], in interviews with students and residents, showed that intense emotional response, the hidden curriculum, role confusion, and professional consequences can limit learning from medical errors. Edmondson (1999)^[23] holds that psychological safety enables team learning behavior, such as discussing failures and seeking feedback. Reason (2000)^[24] shifts error from individual blame toward a systemic reading. NEXO's proportional ICD-10 feedback operates inside that culture: error localized, calibrated, and instructive.

ICD-10 proximity feedback

Conventional education gives binary feedback: right or wrong. The consequence is that the player cannot tell the difference between missing by a little and missing by a lot. NEXO uses the WHO official ICD-10 hierarchy (chapter, group, category, subcategory, exact code) as five levels of proximity. Mistaking ischemic stroke for hemorrhagic stroke falls within the same I60 to I69 group. Mistaking pneumonia falls in another chapter altogether. The feedback indicates exactly at which level of the tree the error occurred, and that becomes pedagogical fuel.

ICD-10 is not the curriculum of NEXO. It is a ruler for standardization and audit. The reasoning remains clinical: complaint, hypothesis, examination, differential, confirmation, and explanation. ICD helps the system compare answers, measure error proximity, and connect each case to NEXO Core.

Calibration and the fluency illusion

The learner is a poor judge of what they know. Soderstrom and Bjork (2015)^[25], in an integrative review on learning versus performance, show that improvement in performance during practice can be an unreliable index of long-term retention; recognition fluency should not be mistaken for durable retrieval. The systematic review by Thompson and Hughes (2023)^[8], with eight randomized studies in radiology education, observes the same pattern in visual interpretation: students predict performance with optimistic bias. NEXO turns feedback into a calibration ruler. Getting it right at clue 1 and getting it right at clue 6 both count as correct, but they move different indicators. Missing within the same ICD-10 group versus missing in another chapter produce distinct records. Across weeks, the player accumulates a proximity profile that says where they actually fail, not where they think they fail.

Dual-coding in multimedia cases

Paivio (1986)^[16] proposed that information encoded in two simultaneous channels, verbal and imagistic, has superior retention. In NEXO multimedia cases, three channels converge in parallel: the clue text (contextual verbal), the clinical image (imagistic), and the technical legend (interpretive verbal). The production rule requires that the three be complementary, never redundant. The legend never simply describes what the image shows; it adds the mechanism, the clinical implication, the pathophysiological threshold. Without that, dual-coding collapses into redundancy and the gain disappears.

Controlled cognitive load

Sweller (1988)^[17] introduced the concept of cognitive load in learning; Sweller, van Merriënboer, and Paas (1998)^[26] consolidated the tripartition into intrinsic (from the material), extraneous (from the design), and germane load (useful for building mental schema). NEXO works against extraneous load. Maximum of four media per case, distributed across adjacent clues. Fixed render order: text clue first (fastest channel), image next (player-paced), audio only on deliberate action (slowest channel, requires play). No auto-play, no carousel, no tab that hides content, no blocking popup.

Difficulty calibrated by media modality

Adding an image or audio to a case changes the cognitive task, not just the visual. A classic chest X-ray of consolidation simplifies the pneumonia diagnosis: the signal is practically handed over. A subtle dermatoscopy of melanoma in situ does the opposite, forcing the learner to discriminate findings they may have never seen. The same disease, with or without media, is two pedagogically different cases.

NEXO keeps two difficulties per case. The base difficulty (Easy, Medium, Hard) is assigned by the clinical design of the case. The media-modality floor raises the final difficulty when the media content is interpretively expensive. The published difficulty is the maximum of the two.

The floors are opinionated, not administrative. Histopathology demands cell-pattern recognition under the lens and is rarely Easy in playable format: the floor is Medium. A multimodal case with simultaneous image and audio demands synthesis of two distinct channels under time pressure: the floor is Hard. Attaching a CT scan to a case classified as Easy triggers automatic reclassification to Medium. The result is that the learner who selects Easy mode never finds a subtle imaging finding hidden inside: the mode is reliable.

Sinapses as signal, not as payment

Deci, Koestner and Ryan (1999)^[11], in a canonical meta-analysis, showed that extrinsic reward for a previously intrinsically motivated activity produces a drop in intrinsic motivation: the over-justification effect. NEXO was designed to avoid that. Sinapses are not payment for time: they are a signal of competence. A correct answer on the first clue is worth more sinapses than a correct answer on the sixth. Errors cost, but the cost is educational, not punitive. Archive mode, deliberately, has no timer and no visible leaderboard during the session: it is where the player revisits cases from the library at their own pace, with less extrinsic pressure.

Interleaving across specialties

Studying all of cardiology, then all of neurology, then all of pulmonology, is blocked practice. Studying one cardiology case, then one neurology case, then one pulmonology case, without repeating the same room in sequence, is interleaving. The systematic review by Thompson and Hughes (2023)^[8] identified only two randomized studies on interleaving in radiology education; specialty-specific evidence remains scarce. The most solid cognitive basis comes from general literature: Rohrer and Taylor (2007)^[27] showed, in mathematics practice, that interleaving improves discrimination between similar categories compared with blocked practice, and Soderstrom and Bjork (2015)^[25] describe the mechanism as a desirable difficulty; the learner is forced to compare and select the correct rule at each item. NEXO's daily mode runs on interleaving by design. Each case belongs to a specialty distinct from the previous one, and the weekly sequence shows wide rotation. The player switches mental models with every case, instead of applying the same pattern. Uncomfortable in the short term, advantageous in the long.

Lim and Veasuvalingam (2025)^[28], in a mixed-methods study on online case-based learning among medical students, report positive acceptance and perceived gain in clinical reasoning. The relevant point for NEXO is structural: cases must allow hypothesis exploration, discussion, and feedback on the reasoning. Those criteria sit at the spine of NEXO: 6 progressive clues and proportional ICD-10 feedback.

Clinical Governance and Scientific Audit

The initial generation of a case is only the first draft. What defines the educational quality of NEXO is the audit that follows.

Each case is confronted with NEXO Core, our medical base structured around clinical concepts, reference chapters, ICD codes, and links between related topics. This corpus functions as a verification layer: the system does not only assess whether the text looks correct, but whether the diagnosis, the clues, the differentials, the explanation, the ICD, and the translations stand up against organized medical sources.

NEXO Core works as the brain of the system, organized as a living clinical library. Inside it, every disease, syndrome, examination, physical finding, and ICD-10 code occupies its own page, and each page is stitched to the others through the relations clinical reasoning recognizes: pneumonia talks to its etiological agent, to its radiological pattern, to its differentials, to the reference chapter that describes it. The result is a dense web, where touching one concept lights up dozens of others that surround it.

When a case is audited, our agents do not read linearly. They enter that web, open the concept of the proposed diagnosis, walk through the neighboring concepts, check whether the case clues appear among them, verify whether the reference chapter supports the explanation. It is a structural sweep against an organized body of medical knowledge, not loose interpretation of text. This allows detection of three types of problem that a superficial review usually misses: clinical inconsistency, poorly anchored diagnosis, and pedagogically weak explanation.

Current coverage of NEXO Core

The library currently covers concepts from the following specialties:

• Cardiology
• Pulmonology
• Neurology
• Gastroenterology
• Endocrinology
• Rheumatology
• Hematology
• Nephrology
• Clinical oncology
• Infectious diseases
• Clinical toxicology
• Pediatrics
• Medical genetics
• Obstetrics
• Gynecology
• Psychiatry
• Dermatology
• Emergency medicine
• Critical care medicine
• Clinical semiology
• General internal medicine

New specialties are incorporated in planned cycles, and each incorporation re-audits the case catalog automatically.

How the score is decided

The final score of a case is not an average. It is an approval rule by dimensions. A case that is good in translation but weak in clinical integrity does not pass. A case that is clinically correct but contains a clue that gives away the diagnosis too early also does not pass. Curation requires convergence.

The dimensions evaluated include:

Minimum approval requires a high score in every relevant dimension. If the resolver flags low confidence, if the ICD does not appear in the expected concepts, if the related chapters do not support the diagnosis, or if there is clinical inversion in a critical translation, the case is not treated as approved. It returns to revision, correction, or corpus expansion.

The score does not appear to the player. It works only as a publication gate. Every case available in the app has already met the criterion, and that is why the player experience is uniform: there is never a case of doubtful quality on display, and there is no quality badge to present, because quality is a precondition, not an outcome.

This layer is what differentiates NEXO from a question bank generated by AI. The objective is not to produce content quickly. It is to produce playable, short, multilingual, and clinically defensible content.

Acknowledged risks and how we mitigate them

Listing what can go wrong is part of the method.

Over-justification effect (Deci 1999^[11], Lepper 1973^[29])

Rewarding an activity that was previously intrinsically motivated can reduce intrinsic motivation. Mitigation: sinapses signal competence, not time spent. A correct answer on the first clue pays more than on the sixth. Archive mode allows revisiting cases from the library at the player's own pace, outside the competitive pressure of the daily case.

Attention-shift effect (Reeve and Deci 1996^[14])

Under competition, focus can shift from learning to winning. Four studies in the van Gaalen^[3] review showed increased simulator use under competition, but only one showed gains in learning. Mitigation: archive mode is the explicit counterweight, with no timer and no visible leaderboard. The player chooses when to take pressure on and when not to.

Expertise reversal effect (Kalyuga et al. 2003^[20])

An intervention that helps a novice can hinder an expert. Mitigation: graded difficulty (Easy, Medium, Hard) and progressive continent unlocking pace the curve. Empirical study across distinct populations is future research, not yet performed.

Limits

NEXO is a training tool, not a substitute. The cases are fictional, built from classic presentations described in the literature, and exist to fuel diagnostic reasoning in short study windows. They do not cover the full variability of real practice: patients with multiple comorbidities, social context that changes management, unavailable exams, decisions under shift pressure, physical examination, communication with patients and families, ethical dilemmas, interprofessional teamwork. None of that fits into a five-minute playable case, nor should it. NEXO does not replace supervised clinical rotation, does not replace preceptor guidance, is not a medical reference, and must not guide clinical decisions for real patients. Complete training requires contact with patients, with teams, and with error mediated by human supervision, and that part of the work keeps happening where it always has: on the ward, in the clinic, on call, in the classroom. NEXO was designed to complement those stages, not to compete with them.

Partnerships and collaboration

NEXO was built to be audited from outside. The methodology described above, the specialty coverage, the case approval criteria, and NEXO Core are public on this page precisely so that universities, companies, and institutions can test, extend, and challenge the platform.

Collaborative research on diagnostic reasoning, independent validation in controlled studies, integration with undergraduate or residency curricula, case-pack licensing for internal use, and co-development of specialties not yet covered are open paths by design. The conversation runs directly with the team, with no commercial intermediary.

Contact: scientific@nexo.wiki.br

References

1. Buchem I, Hamelmann H. Microlearning: a strategy for ongoing professional development. eLearning Papers. 2010;21:1-15.
2. De Gagne JC, Park HK, Hall K, Woodward A, Yamane S, Kim SS. Microlearning in Health Professions Education: Scoping Review. JMIR Med Educ. 2019;5(2):e13997.
3. van Gaalen AEJ, Brouwer J, Schonrock-Adema J, Bouwkamp-Timmer T, Jaarsma ADC, Georgiadis JR. Gamification of health professions education: a systematic review. Adv Health Sci Educ. 2021;26(2):683-711.
4. Landers RN. Developing a theory of gamified learning: Linking serious games and gamification of learning. Simul Gaming. 2014;45(6):752-768.
5. Xu M, Luo Y, Zhang Y, Xia R, Qian H, Zou X. Game-based learning in medical education. Front Public Health. 2023;11:1113682.
6. Karpicke JD, Roediger HL. The critical importance of retrieval for learning. Science. 2008;319(5865):966-968.
7. Trumble E, Lodge JM, Mandrusiak A, Forbes R. Systematic review of distributed practice and retrieval practice in health professions education. Adv Health Sci Educ. 2024;29:689-714.
8. Thompson AC, Hughes DR. The effectiveness of spaced learning, interleaving, and retrieval practice in radiology education: a systematic review. J Am Coll Radiol. 2023.
9. Ericsson KA, Krampe RT, Tesch-Romer C. The role of deliberate practice in the acquisition of expert performance. Psychol Rev. 1993;100(3):363-406.
10. Chen PH, Roth H, Galperin-Aizenberg M, Ruutiainen AT, Gefter W, Cook TS. Improving abnormality detection on chest radiography using game-like reinforcement mechanics. Acad Radiol. 2017;24(11):1428-1435.
11. Deci EL, Koestner R, Ryan RM. A meta-analytic review of experiments examining the effects of extrinsic rewards on intrinsic motivation. Psychol Bull. 1999;125(6):627-668.
12. Festinger L. A theory of social comparison processes. Hum Relat. 1954;7(2):117-140.
13. Van Nuland SE, Roach VA, Wilson TD, Belliveau DJ. Head to head: the role of academic competition in undergraduate anatomical education. Anat Sci Educ. 2015;8(5):404-412.
14. Reeve J, Deci EL. Elements of the competitive situation that affect intrinsic motivation. Pers Soc Psychol Bull. 1996;22(1):24-33.
15. Hase S, Kenyon C, eds. Self-Determined Learning: Heutagogy in Action. London: Bloomsbury Academic; 2013.
16. Paivio A. Mental Representations: A Dual Coding Approach. New York: Oxford University Press; 1986.
17. Sweller J. Cognitive load during problem solving: effects on learning. Cogn Sci. 1988;12(2):257-285.
18. Kolb DA, Boyatzis RE, Mainemelis C. Experiential learning theory: previous research and new directions. In: Sternberg RJ, Zhang LF, editors. Perspectives on Cognitive, Learning, and Thinking Styles. Mahwah, NJ: Lawrence Erlbaum; 2001. p. 227-247.
19. Locke EA, Latham GP. Building a practically useful theory of goal setting and task motivation: a 35-year odyssey. Am Psychol. 2002;57(9):705-717.
20. Kalyuga S, Ayres P, Chandler P, Sweller J. The expertise reversal effect. Educ Psychol. 2003;38(1):23-31.
21. Serra MJ, Kaminske AN, Nebel C, Coppola KM. The use of retrieval practice in the health professions: a state-of-the-art review. Behav Sci. 2025.
22. Fischer MA, Mazor KM, Baril J, Alper E, DeMarco D, Pugnaire M. Learning from mistakes: factors that influence how students and residents learn from medical errors. J Gen Intern Med. 2006;21(5):419-423.
23. Edmondson AC. Psychological safety and learning behavior in work teams. Adm Sci Q. 1999;44(2):350-383.
24. Reason J. Human error: models and management. BMJ. 2000;320(7237):768-770.
25. Soderstrom NC, Bjork RA. Learning versus performance: an integrative review. Perspect Psychol Sci. 2015;10(2):176-199.
26. Sweller J, van Merriënboer JJG, Paas FGWC. Cognitive architecture and instructional design. Educ Psychol Rev. 1998;10(3):251-296.
27. Rohrer D, Taylor K. The shuffling of mathematics problems improves learning. Instr Sci. 2007;35(6):481-498.
28. Lim WK, Veasuvalingam B. Does online case-based learning foster clinical reasoning skills in medical students? Future Healthc J. 2025.
29. Lepper MR, Greene D, Nisbett RE. Undermining children's intrinsic interest with extrinsic reward: a test of the "overjustification" hypothesis. J Pers Soc Psychol. 1973;28(1):129-137.