Randelshofer et al.: Fifteen UX Leaders in AAA Studios on Pre-Production Decisions — Fukai Reads

TL;DR

How do UX leaders actually make decisions during game pre-production? Ivana Randelshofer (Ubisoft Düsseldorf) with Joseph Tu, Ville Mäkelä and Lennart E. Nacke from the University of Waterloo HCI Games Group, plus Yifan Cao (HKUST) and Reza Hadi Mogavi (McMaster), conducted 40–81 minute semi-structured interviews with 15 senior / director-level UX professionals from AAA studios (Blizzard, EA DICE, Guerrilla, Larian, Remedy, Ubisoft and others). Recordings were analysed with reflexive thematic analysis — a qualitative tradition that keeps the researcher's interpretive standpoint visible while themes are built up.

The picture that emerges is plain but important. Pre-production decisions blend three logics that shift with context: an academically-grounded approach (Nielsen's heuristics, Fitts' Law, Cognitive Load Theory), an experience-based approach (internal style guides, design pillars) and a gut-feeling-driven approach. Academic frameworks are not carried in raw; they are translated into internal words like "ambitions" so teams can share them. Organisationally, two layers carry the work: cross-functional strike teams focused on features, and competency teams as expertise pools that consult across strike teams.

I read this today to translate its findings back into the language of puzzle and game making. The paper is an arXiv preprint (arXiv:2608.00313, posted August 2026, not peer reviewed) and its sample is 15 managers at AAA studios. Read it as "this is how AAA UX leaders talked about it", not "this is how the industry works".

Introduction

The authors are Ivana Randelshofer (Ubisoft Düsseldorf, Germany), Joseph Tu (University of Waterloo, Canada), Yifan Cao (Hong Kong University of Science and Technology), Reza Hadi Mogavi (McMaster University, Canada), Ville Mäkelä (University of Waterloo) and the senior author Lennart E. Nacke, who leads the HCI Games Group at Waterloo. The team blends industry and academia; the first author works inside an AAA UX department. arXiv ID 2608.00313, category cs.HC, submitted in the July–August 2026 window. As of the first version I could see, no peer-reviewed venue is indicated, so I treat it as a preprint.

I chose this paper for a slightly different reason than usual. Even in a small operation like Puzzlebyrinth, we decide every day how a board should be shaped — sometimes with reference to academic difficulty models, sometimes with acquired reflex, sometimes with pure feel. I wanted to see how veterans working inside massive AAA teams talk about moving between those three sources.

This is not an AI or PCG paper but a qualitative field study. There are no equations; instead fragments of the fifteen interviews carry the theoretical spine. Quotations arrive labelled "Expert 12 said…", which makes it easy to translate their situation into your own studio while reading.

Background

HCI and games research have piled up UX frameworks. Nielsen's usability heuristics (a list of rules of thumb for evaluating interface usability, codified in the 1990s), Fitts' Law (an old motor-control law relating movement time to target size and distance) and Cognitive Load Theory (an educational-psychology theory about the relationship between working memory capacity and the load placed on it) are typical examples. They look clean on paper.

But once carried into real production they routinely fail to click, and this has been said inside the industry for years. Part of the reason is straightforward: the frameworks speak only of the end user's experience, while AAA production sits inside commercial viability, platform constraints, team structure, deadlines and publisher contracts. Bridging that gap is a shared challenge for most applied research, UX included.

What makes this paper worth reading is that it approaches the gap from the industry side rather than the academic side. Instead of "let us re-translate our frameworks for practitioners", it asks "what do you actually do — in your own words?" There are single-studio UX reports and surveys of game-jam participants in the literature. A qualitative study targeted specifically at senior AAA UX leaders, analysed with reflexive thematic analysis on 15 interviews, is uncommon in what I have read.

Approach / Method

The design is a semi-structured interview study (an interview format where only the topic skeleton is prepared in advance, and the details follow the conversation). The sample is 15 UX leads / seniors / directors from AAA studios. Recruitment was via social media and snowball sampling, with a screening questionnaire verifying seniority and function. Studios named as represented in the sample include Blizzard Entertainment, EA DICE, Guerrilla Games, Larian Studios, Remedy Entertainment and Ubisoft, though quotes are anonymised so a given quote cannot be tied to a specific studio.

All interviews were conducted online, audio-only, by the first author, running 40–81 minutes each; pre- and post-interview reflexive notes were kept to make the researcher's position visible. Analysis is reflexive thematic analysis. The first two transcripts were coded independently by all three authors as an exploratory open pass, then reconciled across three collaborative dialogue rounds while building a living codebook, after which the remaining transcripts were coded together. Themes were organised on Miro using an affinity-clustering approach (grouping similar utterances physically close to each other).

The work was approved by the University of Waterloo REB (#45589) and no monetary compensation was offered. Epistemologically the authors take a constructivist stance (reality is co-constructed through the accounts told) and explicitly frame themes as constructed rather than "discovered". This choice is consistent with their later argument that frameworks should be discussion starters, not prescriptions.

Findings

Theme 1: three co-existing decision logics. Academically-grounded work rests on Nielsen's heuristics, Fitts' Law and Cognitive Load Theory, but Expert 9 remarks that "frameworks often neglect the commercial and organizational realities of game development". Expert 12 translates Nielsen's heuristics into an internal word — "ambitions" — precisely to build "a shared understanding, a language". Experience-based work is codified in internal style guides and project-specific documents; Expert 8: "we provide them with documentation saying if you build an interface for the keyboard and mouse … these are the rules you should consider." Gut-feeling-driven work is spoken about openly; Expert 3 says decisions get made over "late-night beers" and conversations with key stakeholders.

Theme 2: the organisational structures that carry those three logics. Strike teams gather cross-functional members (designers, programmers, QA, narrative) around a specific feature or problem; Expert 15 describes a "very well-structured communication structure" enabling fast iteration. Competency teams pool expertise (UI, UX research etc.) and advise across strike teams, keeping consistency; Expert 6 says "friction is a good thing" in context, valuing necessary conflict rather than trying to eliminate all of it.

Theme 3: why research does not reach the floor. On time, Expert 5 says "the train is rolling" — once production begins, calm redesign is off the table and only "bare minimum solutions" fit. On visibility, Expert 12 prefers conference talks that show "experience and time and application" over academic papers. On vocabulary, Expert 2 notes that unfamiliar theoretical language can "stop the discussion" with stakeholders who lack the shared terms.

Theme 4: shared language and collaboration tooling. Design systems — internal foundations bundling typography, element sizes, console support conventions and so on — are described by Expert 15 as solving "a lot of problems" and cutting exploration time. At the same time Expert 13 says "games are such individual pieces of software that it is often hard to work with generalized information", asking for frameworks shaped like modular "conceptual Lego pieces". The authors then reassemble these into a proposed Model of Adaptive Design Judgment (ADJ).

Use cases

First, explicitly reserve a "window for theory" in pre-production. The paper's recurring lesson is that once production starts, theory rarely lands. If you are planning a new puzzle-game mode, carve out a one- to two-week block up front for reading, citation and hypothesis-writing, and leave the frameworks you consulted as footnotes on the concept doc. Later you can reconstruct why a decision was made.

Second, treat translation of frameworks into internal shared language as a deliberate task. Expert 12's translation of Nielsen's heuristics into "ambitions" is a small but symbolic move. For a hyper-casual puzzle, rewrite Cognitive Load Theory as "how many elements the opening UI asks the player to hold in mind at once" — a working-language sentence, not a definition. Pinning that translation in Slack or on a single Notion page reduces friction later. As Expert 2 warns, unfamiliar theoretical terms can "stop the discussion" outright.

Third, build a small-team equivalent of strike / competency teams as two role layers, not two headcounts. Even with the same two or three people, rotate between "feature units" (one puzzle mode at a time) and "competency units" (board generation, difficulty modelling, localisation, accessibility). Marking a meeting as "today is strike" or "today is competency" keeps agendas from bleeding into each other and preserves what Expert 6 calls the "necessary friction".

Fourth, build a small design system even for a puzzle catalogue. Card frame sizes, board colours, font ramps, console-specific button assignments — anything decided again and again should live in a document. Expert 15 says this "reduces exploration time". But as Expert 13 cautions, binding each puzzle too tightly to the system kills its individuality; I would write the doc as "strong principles, permitted exceptions".

Fifth — and to me this was the most useful implication — mark every design decision in the plan as "theory", "experience" or "instinct". Once you accept that three sources coexist, colour-coding them makes it possible to revisit only the instinct-driven decisions later. For a daily-puzzle operation where decisions accumulate every day, this visibility drops the monthly review cost sharply.

Limitations

The authors themselves list three limitations. First, the sample is restricted to AAA studios and does not generalise to indies or mid-size studios, whose constraints differ fundamentally. Second, because studio identity is anonymised, individual quotes cannot be tied to a studio's culture or organisational structure, so context-dependent variation is partly obscured. Third, the data are a cross-sectional snapshot and do not follow decision-making longitudinally across a project's life. The same person can speak differently at different stages.

The authors also note that industry-wide layoffs occurred during data collection (they cite the IGDA Developer Satisfaction Survey), and honestly acknowledge they cannot know how that background affected candour on stakeholder pressure and production reality.

Where I add my own observations as Fukai: 15 participants is a respectable size for reflexive thematic analysis, but qualitative work should be judged on the richness of insight, not frequency. If a theme (such as "instinct") leans heavily on one or two speakers, presenting it as a general picture risks over-generalisation. As far as I can tell, the instinct theme is strongly carried by Expert 3's "late-night beers" anecdote, and it is not fully clear from the reported text how much other participants echoed it.

Recruitment via social media and snowball sampling also biases the sample toward practitioners who take UX seriously and are willing to speak. Studios that treat UX lightly are unlikely to be represented. A reader should take this as "how AAA UX leaders spoke of it" rather than as an industry-wide fact, and I would resist categorical claims that begin with "the industry does…".

How Fukai Reads It

Here is my own reading. I want to place this paper in the lineage of Schön's reflective practitioner — the old claim that experts are not deducing from principles but being called back by the situation while objectifying their own action — landed for the modern game industry. The Model of Adaptive Design Judgment (ADJ) the authors propose is not a model for choosing which of the three logics is right; it accepts that all three run at once and asks where in the organisation the blending should sit. In the vocabulary of design criticism, that is closer to turning the design process itself into a design system. I would keep this paper visible on my desk as a daily reminder that, in puzzle making too, decisions can come from three sources and I should be honest about which one drove each call.

Closing

If you want a map, first read the qualitative work coming out of the HCI Games Group over the last decade — Nacke and colleagues on player experience and engagement, and their studies of collaborative design in game jams — and you will see the lineage this paper carries. Widen a step and place it against Donald Schön's The Reflective Practitioner (1983): the question this paper touches, how experts decide in situ, is much older and much wider than games UX.

From the practitioner side, I would recommend reading it alongside a book like Celia Hodent's The Gamer's Brain. Practitioner books explain the floor from the framework side; this paper explains the frameworks from the floor side. Squeezing pre-production judgment between the two lets the complexity of it feel a little more organised.

References

Papers and resources referenced in this article:

・Theory, Experience, and Instinct: A Glimpse Into AAA Game Processes and How UX Leaders Navigate Pre-Production (Ivana Randelshofer, Joseph Tu, Yifan Cao, Reza Hadi Mogavi, Ville Mäkelä, Lennart E. Nacke, 2026, arXiv preprint arXiv:2608.00313)

・HTML version of the same paper (main text, figures, appendices)

・DOI: 10.48550/arXiv.2608.00313 (this is the arXiv-issued DOI, not a peer-reviewed venue DOI)

・HCI Games Group (University of Waterloo, Lennart E. Nacke's lab) (the senior author's lab)

・Related books: Donald A. Schön, The Reflective Practitioner (Basic Books, 1983); Celia Hodent, The Gamer's Brain (CRC Press, 2017)