Self-Determination Theory and HCI Games Research: Unfulfilled Promises and Unquestioned Paradigms

Ryan and Rigby posit that “the impact of nearly every element of game design can be seen as a function of its relations to basic needs” [491, pp. 167–168]. The need for competence, for example, is said to be satisfied in scenes of optimal challenge and by three types of feedback that operate at different temporal scales [445]: granular feedback, the immediate feedback players receive for each action, such as blood that appears in response to shooting an enemy; sustained feedback, which games use to recognise and reward “your skill and ability at being consistent” [445, p. 24], such as note streaks in Guitar Hero [229]; and cumulative feedback that “recognizes the player’s skill and accomplishments in ways that persist” [445, p. 29] over the lifetime of the game, such as overall progress. Autonomy, in contrast, is said to be satisfied by games that allow players to enact aspects of their identity—or their avatar’s—via self-expression; the degree of perceived choice in what to pursue in the game, as in RPG sidequest design; and the degree of volition games support through the provision of narrative meaning, structure and rationales for player behaviour [445]. Finally, games are argued to satisfy relatedness primarily through a sense that the player(-character) matters to the game world, whereby “great single-player experiences with NPCs […] succeed with players partly because they provide thoughtful contingent reactions that successfully yield relatedness satisfactions” [445, p. 71]. In multiplayer modes, Rigby and Ryan suggest that relatedness is satisfied by mutual support to achieve objectives, or via good-natured, ‘constructive competition’ in which players display sporting conduct. Notably, while these claims about need satisfaction could produce a deeper understanding of how game design can influence PX, all of the above posited causal relations remain empirically untested in published SDT literature.

The final two aspects of the PENS model were contrived entirely for games research and do not otherwise appear in SDT: first, the notion of intuitive controls (sometimes ‘mastery-of-controls’ [424]) refers to the extent to which players feel comfortable learning and becoming proficient at a game’s control scheme [492]. Specifically, intuitive controls have been theorised as the “price of admission” [492, p. 351], that is, a necessary, albeit not sufficient condition for experiencing autonomy, competence and immersion; the relation to relatedness remains unspecified. In practice, the intuitive controls factor has primarily been used to predict [e.g., 492] and manipulate competence satisfaction [424]. Moreover, results from regression analyses suggest that the effect of intuitive controls on intrinsic motivation is mediated by competence [424, 492] and—occasionally—autonomy [492]. Conversely, a lack of intuitive controls has been found to indirectly promote aggressive thoughts and affect by means of impeding competence [424].

The final PENS concept is immersion (sometimes ‘presence’ [492] or ‘immersion/presence’ [444]), which SDT scholars define as “the illusion of non-mediation” [492, p. 350], a perspective they draw from prior conceptual work outside SDT [see 341]. More specifically, SDT offers a tripartite model of immersion that decentres audiovisual quality and instead corresponds to players’ sense of being physically, narratively and emotionally present in the game world. However, only physical presence has a clear basis in prior scholarship (i.e., presence as transportation [341]). The origins of emotional and narrative presence, and their relevance to immersion as defined in SDT games scholarship (i.e., “the illusion of non-mediation” [341]), are not specified. Likewise, immersion’s importance for SDT has never been fully explicated—its purported relationships with other theoretical concepts remain vague (e.g., “we believe intrinsic motivation in gaming contexts is associated with presence” [492], p. 350). This indeterminacy transfers into how immersion has alternatively been used as a dependent variable, predicted by need satisfaction [426, 492], and as a moderator [425, 428] that purportedly “magnifies the effects of experiences in virtual contexts on behavior in the real world” [428, p. 70]. We also note that some SDT research [513] indicates that mindful awareness—a key facet of wellbeing in SDT (see Section 3.4)—is inhibited by immersion,⁵ suggesting that this aspect of SDT games scholarship is actually inconsistent with the wider theory. Although SDT scholars have suggested that immersion and mindful awareness do not inherently conflict [Rigby, personal communication], the empirical discrepancy is yet to be resolved.

The PENS scale was developed to operationalise players’ experience of need satisfaction, immersion and intuitive controls [444], and first formally introduced in [492]. Competence items reflect how capable players feel during play; autonomy items focus on players’ perception of freedom and choice in the game; and relatedness items pertain to the extent to which players feel connected to other players. Due to the scale’s proprietary status, the full list of items has not been made openly available. The most common version appears to be a 21-item scale, where autonomy, competence, relatedness and intuitive controls are measured with three items each, though some studies employed 5-item versions of the autonomy and competence subscales [492]. Notably, the immersion subscale assesses players’ sense of physical, narrative and emotional presence with three items each [492]. However, all known instances in SDT games scholarship combine the 9 items into a single mean immersion score [426, 428, 492]. Moreover, while it has been claimed that the PENS scale was “validated in two rounds of confirmatory factor analysis using survey data from 2,000 regular video game players” [426, p. 246], to our knowledge, this analysis has never been published.

To update our initial corpus of 110 CHI and CHI PLAY papers [559], we searched the Scopus database and ACM Digital Library on 26 November 2021. We searched Scopus for relevant publications from ACM Transactions on Computer-Human Interaction (ToCHI) and International Journal of Human-Computer Studies (IJHCS), as they are considered high-impact venues in HCI [250]. While not unequivocally considered a core HCI publication venue, our search string⁷ additionally revealed Computers in Human Behavior (CHB) as the most prolific publisher of SDT-based games research papers among all venues. Many of these papers are in turn cited in HCI games publications, and several highly cited CHB papers pertain to SDT and gamification [e.g., 366, 493]. We acknowledge that other venues also publish relevant work [e.g., 154, 362, 430, 553] but decided to focus on what we believe are among the most high-impact venues.

We additionally searched the ACM Digital Library⁸ for papers from CHI and CHI PLAY published after our earlier literature review [559, i.e., proceedings published in 2020 and 2021], as well as the Foundations of Digital Games (FDG) proceedings,⁹ a SIGCHI co-sponsored conference covering a wide disciplinary range of games research. FDG proceedings not indexed by the ACM Digital Library were downloaded from each year’s conference website. Initial filtering of these papers was conducted manually, by the first author, by searching the pdfs for the key terms used when searching Scopus and the ACM Digital Library. Searching in this way returned 320 new publications that were not found in our previous review, increasing our sample from 110 to 430 papers (see Figure 3).

Because we were interested in understanding the ways that games research within HCI¹⁰ applies and discusses SDT and its concepts, we excluded all review papers, or any papers with only cursory relevance to either SDT or games (n = 171, see Supplementary Material for the full list). Both authors agreed on all exclusions made on this basis. A paper was categorised as cursorily relevant to SDT according to prior citation analysis guidelines; in particular, when it merely “mentions [some aspect of SDT] in passing, without any context, details, or information” [202, p. 10], or where “[SDT] work is cited in a list, with no further comment or detail on the individual text” [355, p. 831]. For example, the work of Gong et al. [204] was excluded for this reason, as it cites SDT only as part of a list of background literature. When assessing relevance to games, we chose to avoid limiting ourselves to a particular definition of “game” or “play” and accepted any interpretation of “game-ness” made by the paper’s authors. For instance, we included an article about an interactive storybook fitness app [494], on this basis, but excluded another article focusing on persuasive technology more broadly [593]. We also excluded review papers (such as our own [559]) to avoid redundancy. As a consequence of these filtering procedures, n = 149 papers were retained, producing the expanded corpus (N = 259) that formed the basis for our subsequent review.

As with our previous review [559], papers were coded with respect to venue, research domain (e.g., gamification), SDT-related measures, theory use (see next paragraph), SDT-related references, SDT concepts named, concept definitions, observed relations involving SDT concepts, statements about SDT and variations in terminology use. Papers were split between authors, with the first author coding about two thirds of the corpus. Papers co-authored by one of the authors or their (previously affiliated) research groups were coded by the other author, and both independently coded the single paper in the corpus they co-authored [i.e., 554]. The coding spreadsheets are included as Supplementary Material.

In our previous work [559], we broadly coded theory use according to the purposes of HCI theory outlined by Bederson and Shneiderman [40] and Rogers [456]. However, this approach was deemed not sufficiently granular for the aims of the present work, as a theory’s intended purpose does not necessarily reflect its application in the citing work. Instead, we adapted the citation typology by Girouard et al. [202], which was specifically developed for assessing the impact and degree of engagement with theoretical frameworks. Their typology differentiates higher level theory citations (i.e., generative and analytic theory use)—where theory presumably serves as a direct influence on the citing work—from low-level citations that draw from theory to help contextualise the work, but where theoretical tenets are otherwise non-critical (i.e., their omission would not substantially change the work). Unlike Girouard et al. [202], we did not consider individual mentions of SDT as unit of analysis. This would have complicated interpretation due to many papers in our corpus featuring multiple SDT-based references, often in the same sentence. Rather, each paper in the corpus was assigned a primary theory use category (i.e., descriptive, analytic, or generative, see Table 2), reflective of its presumed application of SDT. Papers matching several primary theory use categories (e.g., if they featured both descriptive and analytical SDT citations) were assigned the higher level category.

We further divided primary categories of theory use into multiple secondary categories synthesised from prior citation analysis schemata [61, 202, 355]. For instance, we distinguish ‘analytic–rationalising’ theory use, where SDT concepts and measures are purposefully employed to examine theoretical tenets, from ‘descriptive–methodological’ citations, to account for the prevalence of HCI games papers that employ SDT-based methods without expressing theory-based rationale [559]. All 259 papers, including those from our original corpus [559], were (re)coded in this way, with regular meetings to discuss papers and iterate coding categories. Disagreements during coding were minor; all were resolved in agreement. For example, we decided to add a ‘generative–extending’ category to account for our interest in work that ‘talks back’ to SDT. Given the emphasis on hypothesis testing in SDT games scholarship [445], papers grouped into the analytic or generative theory use categories were additionally reviewed for SDT-related hypotheses, that is, hypotheses derived from SDT or pertaining to SDT constructs.

While the bulk of papers in our corpus made limited use of theory, a fair amount of works (n = 88) explicitly deployed SDT as the cornerstone of the research, with literature reviews that tended to emphasise relevant theoretical tenets and concepts, which informed hypotheses (n = 44) or research questions. These could be relatively straightforward, as in Kim et al., who predicted that (among other things) “feelings of autonomy will be positively related to enjoyment” [303, p. 696]. More elaborate hypothesising is found in work by Smeddinck et al. [517, p. 5597]. After introducing some key tenets of SDT and flow theory, the authors posited that the provision of embedded difficulty choices would increase players’ sense of autonomy and immersion, compared to conditions where gameplay was interrupted by non-diegetic game elements (i.e., a menu prompting difficulty choice). In alignment with the study’s theoretical grounding, the dependent variables were operationalised via the PENS autonomy and immersion subscales.

Other works linked their hypotheses explicitly to SDT mini-theories [e.g., 19, 77, 371, 386], for instance, drawing from CET and COT to theorise about the extent to which gamification elements affect competence: “Given the assumption that points, levels and leaderboards afford competence need satisfaction […] the effect should be more pronounced for levels and leaderboards, since they provide more performance feedback than points only [… further,] causality orientation may moderate the effects of feedback” [366, p. 528]. Beyond the derivation of hypotheses, SDT was also used to post hoc justify results consistent with its theoretical tenets [e.g., 203, 215, 336]. For example, in an OIT-informed study of pro-environmental behaviour, Van Houdt et al. [576, p. 250] conclude that their study “lends support to the hierarchical model of SDT […] that someone’s personality or user type, as for example defined by the Hexad, […] may not be sufficient to predict gamification preferences.”

A few works resorted to SDT to inform their research methodology. Unlike papers in the descriptive–methodological category, these works purposively applied the theory to “understand(ing) day-to-day play experiences in terms of existing SDT concepts” [555, p. 8]. Saksono et al., for example, employed relatedness need satisfaction as a lens for interpreting interview data around the use of a playful family fitness app [494]. Tyack and Wyeth, investigating the role of autonomy need satisfaction for short-term well-being [555], detailed how their experimental manipulation was tailored according to SDT tenets to deliberately thwart autonomy: “As SDT indicates that ‘control […] can be done to people by contextual events and is therefore more easily evidenced [than autonomy support]’ [137, p. 1027], the design process largely focused on […] the presence or absence of pressuring directives, pre-determined puzzle solutions, and elements that limit choice over in-game behaviour” [555, p. 7]. In another example, Tyack et al. outlined how SDT informed selection of a game intended to “consistently support competence regardless of player skill, while providing fewer opportunities to satisfy other basic needs. Previous SDT research [476, 492] identified game controls, progression structure, and genre as particularly relevant” [557, p. 4].

Although papers in the rationalising category generally applied SDT in substantial ways, they also frequently contained unorthodox theoretical claims and methodological practices. Luu and Narayan, for example, critiqued a “meta-analysis [that] included studies in which broad measures of motivation, interest, engagement, and/or even attitudes towards tasks were used as measures for motivation in serious games” [347, pp. 111–112]. In their own study, however, they combined all 21 items of the PENS scale (i.e., which index need satisfaction, immersion and intuitive controls—not intrinsic motivation) to assess intrinsic motivation, engaging in the same questionable research practice.

SDT concepts were also (mis)characterised in somewhat unusual ways. In one instance, a study of young women’s play habits and undergraduate degree selection [251] includes a number of spurious connections between basic needs and other variables, such as self-concept (competence), income deprivation (autonomy) and time spent playing videogames (relatedness). Similar conceptual flexibility is demonstrated in a case study of frustration in ‘casual games’ [453], which links competence to the illusion of control, and suggests that relatedness is satisfied by competition, “even though players cannot really compete, the illusion of control bias lets them believe they can” [453, p. 5].

Finally, the presence of theory-based hypotheses in a paper did not preclude descriptive use of some SDT concepts. For instance, although the aforementioned study by Smeddinck et al. [517] is exemplary for how transparently it grounds its hypotheses in SDT, the same study also included PENS measures for competence, relatedness, intuitive controls, as well as the IMI interest/enjoyment and effort/importance subscales, with no apparent theoretical rationale other than “as additional dimensions rooted in SDT to augment the PENS results” [517, p. 5599].

Despite the fair number of papers making use of SDT to inform their studies and explain their data patterns, we observed comparatively little interest (n = 29) in discussing findings in terms of their implications for SDT, beyond a common acknowledgement that results were (largely) consistent with the theory. Indeed, across theory use categories, results that did not align with SDT were consistently blamed on other factors—study design [e.g., 493], quirks in the data [e.g., 419], or by recourse to other (sometimes previously unmentioned) work [e.g., 167]. Mekler et al. [366], for example, tested SDT-derived hypotheses that gamification would increase competence and intrinsic motivation in an image annotation task—for which they did not find support. They follow this up with a lengthy discussion of SDT-conforming explanations under which their hypotheses might nevertheless hold true, e.g., “the motivational appeal of many games lies in their ability to provide players with challenges to master, hence allowing them to experience feelings of competence. The image annotation task, on the other hand, could hardly be considered challenging” [366, p. 532].

Of course, it is informative to first consider methodological confounds and how divergent results might be explained within SDT. What is striking about the aforementioned examples, however, is the apparent reluctance to consider explanations that may point to potential gaps or incongruities in the theory—and could drive further theory development and specification. Correspondingly, we observed only few instances that more directly contrasted their findings with SDT. Tyack and Wyeth [555], for example, discuss how “the observed variation in emotional trajectories (H2) speaks to the difficulty of characterising these experiences as unequivocally positive or negative […] this interpretation is complicated by their respective associations with need satisfaction and frustration (H3)—which are identified as unequivocally positive and negative in SDT” [555, p. 18].

Finally, some papers placed SDT in conversation with other theories [e.g., 77, 371]. Liao et al. [338], for example, fruitfully combined aspects of SDT with self-affirmation theory to examine links between real-world need satisfaction, self-esteem and future play intention. Other papers in this category discuss multiple theories to make wider conceptual points, as in Sweetser and Aitchison’s work on (extrinsic) reward-based systems and intrinsic motivation [532], which compares perspectives from behaviourism, SDT and flow theory. However, it is not always clear why authors have decided to combine SDT with other theories. A notable exception is work by Poeller et al. that combines SDT and Motive Disposition Theory (MDT) [414, 415], “because SDT […] does not focus on individual need strength. Therefore, we give an example on how different motivational theories have greater explanatory power when used together” [415, p. 14].

Finally, we observed few papers in our corpus (n = 6) that contested aspects of SDT, and even these tended to be rather indirect. For example, investigating gamified achievements, Groening and Binnewies note that they “do not want to get into the controversial debate if tangible rewards undermine motivation,” but regardless “want to emphasize that we believe digital achievements may be more than just goals with rewards” [215, p. 162]. Our own work [554] is also relatively circumspect, in that it largely serves to critique SDT’s exceptionalist characterisation of videogames by drawing attention to Rigby and Ryan’s claim [445, p. 109] that play can be “comparable to recreational drug use” [554, p. 2].

However, work by Palomäki et al. is notable for its more explicit disagreement with SDT; specifically, the authors interpret their study results as complicating the relationship between competence and intrinsic motivation, arguing that “as one gains experience in an activity, expectations rise accordingly [… which] will most strongly determine the (dis)satisfaction one derives from the task, rather than one’s absolute level of perceived competence” [402, p. 10]. Other instances of more express critique include questioning the “increasingly prescriptive links between need satisfaction and specific game elements” made in SDT games scholarship [555, p. 19], or challenging the explanatory capacity of SDT, e.g., “the need for relatedness is a universal need that all individuals share. Yet, SDT—or its myriad mini-theories—provides little explanation of how people prefer to satisfy their universal need to relate to others” [415, p. 2].

Only five recorded efforts aimed to extend SDT in some way [i.e., 153, 159, 318, 398, 544]. Dindar and Akbulut [159], for example, attempted to map SDT’s basic needs onto Yee’s motivational framework [615], arguing that “even though these motivational needs are summarised in previous works […] interrelationships among these motivations in video-gaming environments have not been described yet,” and that their study results “can guide further empirical works to scrutinise the interrelationships among these basic needs” [159, p. 124]. Koulouris et al. [318] studied avatar identification in a VR exergame, deploying SDT as one of many theoretical lenses through which to explain the study results. In doing so, the authors also draw from other exergame concepts as well as Jungian psychology to propose two new SDT-related terms: “intrinsic identification, which is based on a recognition of one’s true Self and fostering a sense of oneness with the Self, and extrinsic identification, which is based on personas projected by the ego and drawing the attention away from the Self” [318, p. 10; emphasis in original].

Although these papers show sufficient investment in SDT to propose theoretical extensions—or perhaps as a consequence of doing so—both works contain claims about autonomy that seem to bear little relation to its characterisation in SDT. When linking basic needs to Yee’s framework, Dindar and Akbulut argue (citing Rigby and Ryan [445]) that because “video-game players can form an idealized view of their personalities and have novel experiences which are somehow impossible to find in their real lives,” that “the immersion component of Yee’s framework can be considered equivalent [to] the autonomy need” [159, p. 121]. However, the cited reference does not actually discuss¹² the notion of players’ ideal self, nor specifically link immersion to autonomy need satisfaction. Koulouris et al. instead refer to SDT when discussing their study results, suggesting that “realistic avatars better fulfill the three main psychological needs that facilitate self-motivation,” including “autonomy because a realistic avatar frames their efforts as self-competition independent of others” [318, p. 9]—the theoretical basis of which is left unexplained. In contrast, Deterding’s grounded theory of contextual autonomy support in play [153] hews more closely to CET, outlining testable propositions of how the presence or absence of situational demands (e.g., making time for play) impacts autonomy need satisfaction.

Several authors used SDT as a tool for design (n = 16, 6.20%)—both as part of the design process itself, and (usually following a user study) to formulate design implications. The ways that SDT actually informed design in these works was not always entirely clear, however, as in Fatehi et al., who “were inspired in our design approach by considering player motivations” [177, p. 3; emphasis in original]—which here refers to an amalgam of Bartle’s player typology [37], the Hexad model [544] and SDT. The authors’ actual design description makes only passing mention of (an unspecified type of) motivation; for example, in noting that they “tried to characterise the writer and the critic [… to] provide a narrative context that would motivate [participants] for the task” [177, p. 5].

Even when artefact designs were explicitly identified as being ‘informed’ by SDT, however, these connections were often limited to the inclusion of particular game features (e.g., points, badges and leaderboards). In their study of gamified learning systems, for example, van Roy and Zaman note that “by awarding group points we also intended to support students’ feelings of competence, as these points served as reinforcing feedback” [579, p. 41]. The relative terseness of these theoretical links often stood in contrast with more thorough descriptions of the designs as a whole, typically supplemented with arrays of annotated screenshots and (in some cases) a link to the artefact itself.

That said, some works stood out for attempting to more clearly ground their design work in SDT [e.g., 17, 582]. Of particular note is work by Miller et al. [370], who describe a process of ‘theory adaptation’ to translate concepts from SDT (and Cognitive Load Theory) into generalisable game design features, with a view to redesign the citizen science game Foldit. Competence, for instance, was first ‘operationalised’ as “the player’s understanding of the tools available to them” [370, p. 5] and then instantiated as “giving the players more instruction on how to use the tools available to them and less instruction on what actions they should take” [370, p. 6; emphasis added].

Recognising that theoretical propositions may not readily translate into design [191], we had expected papers in this category to demonstrate more flexible applications of SDT—more frequently, however, we observed more standard misinterpretations of the theory. Barata et al., for example, defined autonomy in gamified education as “a sense of control over the learning environment” [32, p. 556], a fairly common misconception that has prompted a number of clarifications within SDT¹³ [e.g., 136]. A similar definition was provided by Tsai et al. (i.e. “autonomy refers to a person’s sense of control over his or her own choices” [550], p. 2), whose work on a gamified social fitness system also misinterprets internalisation (cf. Figure 2) as referring to all forms of (a)motivation “as the internalisation of motivation increases, one’s motivation can change from amotivation to intrinsic motivation” [550, p. 2].

As in the analytic papers, when user studies failed to produce results in line with SDT, the theory itself was not questioned; rather, external factors or the artefact itself were identified as potential causes. After observing somewhat mixed results from their aforementioned design changes, for example, Miller et al. lamented that their “work was built on the existing structure of Foldit (now 11 years old)” [370, p. 9]. Notably, however, they also acknowledged that “the [design] operationalisation process itself is a limitation in that there is a layer of abstraction between the theory and the implementation,” speaking to the issue of theory translation—a concern also shared by industry practitioners.

Most GDC talks employed SDT as a conceptual frame to reason about what makes games engaging and commercially successful. Rigby [440] explicitly framed his presentation in these terms, emphasising that “the underlying hypothesis [of the PENS model] is that you’re really going to have sustained fun and commercial success when you’re satisfying intrinsic motivational needs in your player.” Several presenters similarly focused on need satisfaction with a view to improve profit; for example, “if your microtransaction just satisfies one of these needs, then you’re doing great, and it’s going to have value, and people are going to buy it” [118].

Among SDT concepts, autonomy was most prominent, with a total of 54 mentions (including eight mentions of “agency” in [257]). Notably, Rigby referenced (to our knowledge) unpublished study findings that “autonomy actually has the strongest relationship of all the needs to sustained subscriptions in MMOs” [440]. He further clarified that autonomy “isn’t exactly freedom […] you can actually constrain people’s choices as long as whatever path you’ve put in front of them, they feel that they’ve chosen to take,” emphasising its basis in an internal locus of causality. Other presenters, however, primarily framed autonomy in terms of “having options” [78], “meaningful choice and self-expression” [244] and making “decisions that matter” [257]—terms that mirror those used in videogame marketing [396].

Besides need satisfaction and the PENS model, intrinsic motivation was also considered important. Shirinian [515], for instance, emphasised that “if your game intrinsically motivates your players, they will want to play more often, they will be more deeply engaged […] they will evangelise your product more, and in fact they’re also learning themselves”. Others considered both intrinsic and extrinsic motivation core to the PX. Hodent [244], for example, indicated that “intrinsic motivation is even more powerful [than extrinsic motivation] for long-term engagement,” but “players’ engagement is typically driven by extrinsic motivation, because they want to achieve external rewards, and avoid the opposite, typically the absence of rewards.” Similarly, some [245, 333] speakers discussed ‘non-contingent’ (or in strict SDT terms, ‘engagement-contingent’ [132]) rewards. Lewis-Evans [333] described non-contingent rewards as “participation awards” that “are less likely to be seen as controlling […] because you don’t have to do anything to get them.” Drawing from design examples from racing games (rubber-banding), MOBAs (passive resource generation) and ‘casual’ games (catchup mechanics), Lewis-Evans further argued that “non-contingent rewards can be quite useful for […] balance, and making sure people experience more of your game.”

Rigby [440] described the PENS model as “a divining rod” to assess “how is this [feature] going to satisfy the player?” Indeed, we observed a number of instances where speakers drew on SDT concepts to analyse particular game designs. Hudson [257], for instance, explained the effectiveness of Red Dead Redemption’s [449] climactic shootout scene through the player’s learned competence. In particular, the analysis focused on the game’s Deadeye mechanic, which provides brief invulnerability and slows time while the player selects targets. Hudson notes that “as you play, you’ve probably gone into Deadeye hundreds of times… if you were really good, you could maybe peel off five guys before they got to you.” Hence, in the final scene, when “the game forced you into Deadeye, and you’re surrounded by twenty people, you panicked […] you said ‘oh my god, my competence as a player, this isn’t enough’ […] They sold that sense of despair and finality, through using that mechanic, better than any cutscene could have.”

Burnell instead used autonomy frustration to make a point about designing choices, contrasting “a scene [in Portal] [574] where you have to destroy an inanimate box” with a similarly conflicting scene in Portal 2 [575]. Noting that “people love Portal’s scene, and hate Portal 2’s scene,” Burnell observed that “in both […] you have your autonomy taken away, but in [Portal], the villain is making you do this heinous act, and in [Portal 2], the game designer’s making you do it […] in one case, the player is angry with the villain, and in the other case, the player is angry with the game designer.”

We identified some cases where SDT concepts were used to develop a folk theory [192]; in other words, where SDT concepts were adapted into developers’ own conceptions of PX and good game design. As with other instances where psychological theory has been applied to game design practice (e.g., the big five personality model in [583]), these folk theories vary in terms of how faithfully they reflect SDT tenets. Griesemer [213], for example, described Halo’s [75] dodging and cover mechanics in terms of competence and autonomy, respectively, to emphasise the contextual value of need satisfaction, claiming that “competence in a shooter, specifically dodging, is much more fun when you’re under low pressure, and autonomy is much more fun when you’re trying to tackle a problem that’s much more difficult.” In this way, he argued that “when games satisfy these complex needs, that’s when they’re really fun. And this complexity comes from the context that the needs are being satisfied in.”

Lewis-Evans [333] instead discussed common industry practices for supporting intrinsic motivation, noting that “a lot of companies […] talk about intrinsic motivation [in terms of] goal structures, and adding a bunch of stuff on top of the game.” Acknowledging the merits of such approaches, Lewis-Evans championed good game feel [535] as “the ultimate” factor. In this view, “games that are commonly cited [as intrinsically motivating] like Destiny [76] […] just feel satisfying to play. The action of just moving a controller or a mouse feels impactful and immersive.”

Finally, we noted a few instances where game developers critiqued the extent to which need satisfaction could meaningfully speak to game design. Though these critiques were somewhat facile in their engagement with theoretical tenets, they form notable counterpoints to perspectives that leverage SDT as a conceptual frame. Burnell [78], for instance, questioned whether Farmville [624] could really be considered “the pinnacle of game design,” even though “you can plant crops in any pattern you want, giving you autonomy. The only skill is clicking, so no-one feels a lack of competence. And your friends are relying on you to play, and benefit from you playing, giving you relatedness.”

In another example, Griesemer [213] was critical of the relevance of SDT-based study findings for game design practice, with respect to an experiment [426, Study 4] where the researchers “modded Half-Life [573] […] one version […] was [violent] and [the other] was non-violent; it was sort of like a game of tag.” Although some details of the study were misconstrued,¹⁷ his conclusion likely echoes industry sentiments: “we all know that Tag of Duty would not sell.”

The HCI literature has proposed several attributes of ‘good’ theory, for example, precise counterfactual propositions [400] or generative capacity for design [38, 248]. Prior work has also speculated about the factors that inhibit theory use in HCI, such as iterative design practices [400]. Rarely, however, has the literature considered what makes a theory ‘successful’, that is, what conditions facilitate its uptake in a community of practice (e.g., HCI researchers, industry professionals [455, 456])—conditions presumably also advantageous to establishing (a specific) theory as a motor theme [314]. SDT stands out for its widespread adoption across research areas and application domains, including a growing presence in HCI [30]. Yet for all assertions that “SDT stands unrivalled in its popularity among HCI researchers” [417, p. 261], the reasons for this remain at best implicit. Examining SDT’s popularity in games may therefore help elucidate the circumstances under which theory comes to shape research and design practice (e.g., as paradigm), and why sometimes even good theories get relegated to the sidelines. It may also help explain why SDT has been privileged over alternate theoretical perspectives in games research [417, 554]. The following observations should not be understood as causal claims per se, but rather serve to spotlight taken-for-granted aspects of theory use.

Critiquing the primacy of SDT in HCI games research over other motivational frameworks, Poeller and Phillips muse “whether some researchers may be defaulting to SDT because they feel it is the only option available” [417, p. 261]. We argue that this is no coincidence. SDT scholars have actively cultivated interest in the theory through multiple lines of translational work [114]. This is perhaps most evident in the PENS model, which represents a translation of BPNT and (to a lesser extent) CET to games. Compared to other motivational theories that still await such specification (but see work on GameFlow [533, 534]), SDT concepts may appear more intuitive and relevant in the context of games.

With need satisfaction and intrinsic motivation, in particular, SDT put forward a conceptualisation of ‘good’PX that is applicable to many areas of games research (e.g., design, wellbeing, applied games), without necessarily requiring deep engagement with SDT tenets. With regards to HCI theory, Rogers similarly noted that “stand-alone, one-off terms that conjure up what they mean intuitively have been the most widely taken up—even though they are often used much more loosely and in underspecified ways” [456, p. 83]. Arguably, it is precisely this looseness that renders said SDT concepts more intuitive yet prone to misconceptions, more amenable to descriptive applications in research, and which facilitate their adaptation into industry folk theories [e.g., 213, 333].

The PENS and IMI scales have likewise played a major role in popularising SDT in games research, providing a means to operationalise SDT’s conceptualisation of good PX. Indeed, the majority of papers in our corpus deployed SDT-based measures (54.05%, n = 140, see also Table 3), and frequently with no apparent theoretical rationale (n = 76). Although this is emblematic of the often perfunctory treatment of SDT in the literature, it also suggests that these scales readily “graft onto existing practice” [456, p. 84] of HCI games scholars, who rely on self-report instruments for evaluation and to study the PX [275, 581]. In this way, the PENS and IMI questionnaires and the long-standing absence of effective alternatives [329] rendered SDT more accessible to HCI games research, but also helped cultivate the conditions for shallow engagement with the theory. It remains to be seen whether descriptive-methodological citations, where SDT does not otherwise inform the citing work, will decline now that alternate (i.e., non-SDT-based) validated PX questionnaires are openly available [e.g., 581].

It is worth noting that initial SDT games scholarship [492] not only introduced SDT concepts and measures to games research but also functioned to displace competing approaches [e.g., 15, 37, 615]. For instance, showcasing that the PENS model exhibits improved incremental validity [492] over Yee’s gaming motive framework²⁰ [615, 616], as well as implying that the latter “largely reflect(s) the structure and content of current games” [492, p. 348]. These critiques were reiterated and indirectly amplified in subsequent publications [i.e., 425, 445], such as in calls for ‘clear’ and ‘good’ theory to drive ‘real’ hypothesis testing [445, pp. 167–168]. Of course, such statements may chiefly reflect common academic writing practice [201] rather than intent to disparage other frameworks. Indeed, they represent a (somewhat implicit) example of ‘talking back’ to theory [39, see next section]. Nevertheless, they likely helped further cement SDT’s status as a ‘default’ theory [417, p. 261] in games research.

More speculatively, some of these critiques are perhaps also a product of competing commercial interests²¹ to better position Rigby and Ryan’s UX consultancy, Immersyve. Indeed, SDT scholars actively fostered industry engagement via several translational efforts [114], such as GDC talks [e.g., 440, 442], a popular science book [445] and materials [444] made available on the Immersyve website. Importantly, our analysis of GDC talks suggests that these efforts did not simply serve dissemination purposes, but in introducing SDT as a conceptual frame to reason about sustained engagement and profit [e.g., 440] directly catered to practitioner concerns [289]. Industry practitioners, in turn, appear to reference SDT for the presumed rhetorical power [225] it conveys, with presenters perhaps citing the theory to legitimise existing intuitions [e.g., 213, 257]—indeed, the common assumption at GDC is that game designers have already heard of SDT [118, 333]. Moreover, as a professed scientific theory [440, 481, 483] backed by five decades of research, SDT research may appear more legitimate to game developers than comparably nascent HCI games scholarship.

Finally, “the reception of a theory is shaped by the extent to which a theory resonates with the cultural presuppositions of the time” [158, p. 394]. This also applies to SDT games scholarship, where much of its conceptual apparatus—immersion, intuitive controls, heroic frame, but also PENS notions of competence and autonomy that centre mastery and choice—echoes mainstream game design conventions [322, 396]. It is not clear whether this simply reflects SDT scholars’ personal views on games, or whether this is the product of concerted efforts to establish industry relevance. The latter might account for why the hero construct has been incorporated into SDT games scholarship, not based on empirical evidence, but because it is “[w]hat everyone else refers to” [141, p. 35]; whereas other SDT concepts (e.g., functional significance, continuum of extrinsic motivation)—which in principle could fruitfully expand, nuance and complicate mainstream conceptions of game engagement—have yet to be considered.

But note that theories convey normative meanings [193, 198, 623], with potentially far-reaching consequences when applied to technology design: “As its implications and applications are borne out, every theory becomes an ethical or ideological advocate” [193, p. 1354]. By championing immersion and intuitive controls as hallmarks of need-satisfying play, for instance, SDT scholars imbue them with scientific authority. In this way, SDT games scholarship does not simply reinforce dominant design values, but risks marginalising perspectives on worthwhile game design that run counter to SDT-informed notions of good PX [e.g., 503, 554]. A more overtly disconcerting example can be found in Rigby’s recommendation to GDC attendees [440] that “you can actually constrain people’s choices as long as […] they feel that they’ve chosen.” This may not be detrimental to players’ experience; indeed, perceived autonomy was linked to sustained engagement and future play intention [440, 492]. However, it raises questions about the commodification of need satisfaction (see also [519]) and the ethical ramifications of autonomy manipulation [43, 388]—especially when linked to a theory nominally concerned with self-determination.

HCI games research is likewise embroiled in (perpetuating) these cultural and ideological presuppositions. In their study of games’ need-restoring qualities, for example, Tyack and Wyeth [555]—acknowledging their complicity—question whether SDT’s view of wellbeing, as a product of self-realisation [479], over-privileges individual responsibility (see [317, 603]) and “enroll[s] play into practices of responsible self-governance” [555, p. 19]—and correspondingly de-emphasises the state’s role in providing affordable public health care, the ‘safety net’ that SDT otherwise endorses.²²

SDT has been useful for HCI games research [559], it is indeed “popular for a reason” [417, p. 261]. However, this popularity belies a legitimacy trap [165] that has allowed perfunctory applications of the theory to proliferate and lends unwarranted credibility to SDT games scholarship. HCI games researchers cite SDT because it is frequently mentioned in HCI games papers²³, because questionnaires are available, or because industry practitioners reference the theory; rarely to purposively leverage its theoretical tenets. Even scholarship critical of the primacy of SDT in games claims that the theory has been subject to “widespread validation” in HCI [417, p. 261]—a misconception that the present work has hopefully dispelled—and in this way inadvertently perpetuates the very paradigm it intends to challenge.

Continued deployment of SDT as a paradigm impedes HCI games research in several ways: It implicitly guides what phenomena are considered and how they are studied [456], eschewing more reflexive and transparent uses of SDT. It rarely involves ‘talking back’ to theory [39]. It may also narrow the ways we think about games and play [554]. Further, it may conceal theoretical inconsistencies and ideological underpinnings, and crowd out alternate theoretical perspectives—reinforcing use of SDT as an unquestioned paradigm. Lastly, in the absence of translational work in HCI games research, design practitioners are unlikely to notice or benefit from HCI games scholarship.

As indicated by our analysis of GDC presentations, game designers are already familiar with many of SDT’s core concepts. Yet theory is often underspecified for design [191, 455], requiring further translation, and many claims in SDT games scholarship lack empirical support. The analytic and generative uses of theory we observed in our corpus suggest that, in principle, HCI games research is well placed to contribute to game development practice—by testing, refuting or extending theoretical propositions; and translating SDT concepts to design. However, success in this endeavour depends, to some extent,²⁴ on whether HCI games scholars go beyond the currently dominant applications of the theory—namely, perfunctory discussions of need satisfaction and intrinsic motivation (Figure 5)—and further engage with the implications of SDT’s numerous theoretical tenets, and existing findings derived from relevant HCI games research.

One way forward is to investigate unsubstantiated claims in SDT games scholarship to establish links between SDT concepts, game design and behavioural and psychosocial outcomes. That said, most existing propositions (e.g., regarding optimal challenge and competence, see [155]) are insufficiently specified to support hypothesis-testing [499]. Hornbæk [249, p. 188] has identified a parallel issue in usability research, which “contains very few propositions, that is, rarely commits to clear statements regarding the relations among dimensions.” The near total absence of propositions about SDT concepts and games-related phenomena is especially jarring given that the wider SDT literature [e.g., 481] already provides fairly well-articulated propositions. Delivering and investigating such propositions (see Section 9) would facilitate progress in HCI games scholarship, in the form of a cumulative body of research, and allow study findings to ‘talk back’ to SDT.²⁵

What does ‘talking back’ to SDT entail? We apply here Beck and Stolterman’s models of theory use [39, see also Section 10] to compare two examples from our corpus [i.e., 277, 555]. Both papers feature ‘self-determination theory’ as author keywords, suggesting that the theory figures prominently in these works: Johnson et al. [277] explored variations in PX across different game genres and observed that need satisfaction (as measured by the PENS) does not adequately account for the appeal of massive online battle arena (MOBA) games. Their findings are explained in terms of genre-specific properties and that the PENS measure does “not fully capture the components of PX that attract people to play MOBA games” [277, p. 2270]. While the divergent findings present opportunity for ‘talking back’ to theory, the authors leave the incongruity between SDT tenets and MOBA games largely uncommented.²⁶ In this way, the work leaves unclear how (if at all) SDT could inform future work, and (inadvertently) cements its use as paradigm.

Tyack and Wyeth’s investigation of autonomy-supportive play [555], in contrast, leveraged SDT to derive study hypotheses and analyse interview data. The authors ‘talk back’ by comparing their findings to SDT literature, for instance, noting that “it is surprising that the substantial changes observed in autonomy satisfaction did not predict changes in happiness or calmness. The theory itself is somewhat unclear on this point […] identifying the circumstances that produce these ‘typical’ cases […] seems essential to formulate adequate hypotheses” [555, p. 18]. In this way, the work raises open questions for future research and calls attention to opportunities for theory development.

Finally, we found few design applications of SDT in our academic corpus (Table 2), and some authors commented on difficulties [e.g., 370] with translating the theory. Work on intermediary knowledge building in HCI [e.g., 38, 127, 413] may offer some pointers for linking SDT and design in a more principled manner, for instance, by formulating generative questions based in theoretical concepts [38, see also Section 9 for preliminary examples]. These initiatives are worthwhile as a means to further develop theory [114, 529], build on existing HCI games research and improve game development practice. Moreover, our analysis of GDC presentations indicates that industry practitioners’ experiences of applying SDT in design can also create opportunities to (con)test theoretical tenets. Even if these claims are yet to be formally tested via academic studies (e.g. [213], on contextual need satisfaction), they represent serious engagement with theory and its limits, and form fruitful opportunities for further academic investigation [114, 211].

Another promising avenue for bridging theory and practice is to develop translational resources for design [113, 619]—informed by SDT and grounded in relevant HCI games research. In light of the ways that game designers themselves appear to conceptualise SDT in their own work, translating the theory’s tenets into more useful resources for design practice may require some degree of flexibility. This is not to say that we should celebrate work that diverges extensively from SDT tenets, but rather that strict theoretical fidelity may not be the sine qua non of translation [see also 591]. For example, although need frustration is unequivocally considered a negative experience in SDT [470], it may have other applications in PX design e.g., enhancing players’ emotional responses; [78, 257]). Further, translating HCI games research into alternative formats (e.g., online articles, podcasts, GDC presentations [114, 325, 431]) that are more amenable to design practitioners may require shedding some of the formal rigidity that has given rise to structural norms and standards in HCI (games) literature (e.g., see [91, 164]).