Video games invert the expected relationship with failure - rather than avoiding it, players actively pursue games where failure is frequent. Generally, prior work has focused on players encountering hard-coded fail-states as defined by the game (e.g., losing a life). However, scholars have pointed out that game communities often set their own goals (e.g., speedrunners), changing what it means for the player to fail. Further, a more recently explored perspective concerns how developers themselves conceptualize and design fail states. Yet, no known work has explored developers’ perspectives on player experiences of failure within a specific genre. This study contributes to this through semi-structured interviews with 20 puzzle game developers, identifying commonalities in their design philosophies, approaches to designing challenge, conceptualizations of failure, and criteria for puzzle quality. Through inductive thematic analysis, we show that puzzle developers reframe failure away from performance-based definitions. Rather than centering on incorrect attempts, they conceptualize failure primarily as complete player disengagement, viewing it as a failure of the design. Further, puzzle developers regard what some might consider failure as an intentional learning mechanism that encourages experimentation and nudges players toward solutions. Through this paper, we show how developers set player expectations and goals to create a unique relationship with “failure”.

Learning is a complex process that occurs over time. To represent this complex process, interests has been rising in conceptualizing and integrating temporality into model constructions. However, the construction of an event-based temporal model is challenging. Specifically, researchers struggle with translating qualitative heuristics and theoretical hypotheses into quantifiable temporal parameters. Existing methods of parameter derivation also suffer from issues of model transparency and oversimplification of learning contexts. Thus, we proposed a conceptual and methodological framework, Qualitative Parameter Triangulation (QPT), to center human interpretation in model construction. Based on human interpretations, QPT constructs a qualitative loss function and derives temporal parameters using an automatical optimization algorithm. The final step is to check consistency between a global representation with local qualitative evidence given specific learning moments. By presenting a worked example of QPT, we demonstrated the process of maintaining pairwise alignments across interpretation, systematization, and approxi-gation. As a proof of concept, QPT is a feasible framework for determining temporal parameters and constructing event-based temporal models.

Game designers often utilize tutorials to explicitly introduce players to game mechanics. Another design approach is to begin by throwing players into the “shallow end”, introducing them to mechanics through active engagement in easier levels. To compare these different approaches, we created two modified versions of the first 7 levels of the puzzle game Baba Is You: a tutorialized version, and an increased impasse version. Seventy-five players spent up to 20 minutes playing one of the modified versions or the original game and were asked to fill out a survey on their engagement level, perceived difficulty, and understanding of the core game mechanics. Analyses show that players of the tutorialized version made fewer mistakes and rated their version at a lower difficulty than players of the original and increased impasse versions, while all groups scored highly on content knowledge and engagement. This suggests that direct instruction at the start of a game allows players to come to an understanding equal to a learn-by-doing approach more quickly, with fewer mistakes, and without sacrificing engagement. This study does not, however, show that tutorial players are more prepared to take on more challenging content without explicit guidance, a direction proposed for subsequent work.