Gamification in Digital Mental Health Interventions: A Systematic Review of the Engagement–Efficacy–Ethics Trilemma

Abstract

Digital Mental Health Interventions (DMHIs) offer a scalable solution to the global mental health crisis, yet their real-world impact is often hampered by low user engagement. Gamification has been widely adopted as a strategy to enhance adherence, but its implementation creates a complex and often unacknowledged “Engagement–Efficacy–Ethics Trilemma”. This systematic review synthesises the current literature to deconstruct this trilemma, arguing that an uncritical focus on maximising engagement can fail to improve—or may even undermine—clinical efficacy, while simultaneously introducing significant ethical risks. Our analysis reveals a persistent “Engagement–Efficacy Gap”, where increased usage of mobile health applications (mHealth apps) does not consistently translate to better therapeutic outcomes. Furthermore, we map the ethical landscape, identifying potential harms such as manipulation, psychological distress, and privacy violations that arise from persuasive design. The roles of Artificial Intelligence (AI) in personalising these experiences and Human–Computer Interaction (HCI) in mediating user responses are critically examined as key factors that both amplify and potentially mitigate the tensions of the trilemma. The findings indicate a pressing need for a paradigm shift toward an integrated approach that concurrently evaluates engagement, efficacy, and ethical integrity. We conclude by proposing a framework for responsible innovation, emphasising theory-driven design, co-design with users, and prioritising intrinsic motivation to harness the potential of gamified DMHIs safely and effectively. Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, a systematic search was conducted across Scopus, Web of Science, MEDLINE, and PsycINFO for studies published between 2015 and 2025.

1. Introduction

Research on the influence of gamification on engagement, efficacy, and ethics in digital mental health has emerged as a critical area of inquiry due to the increasing prevalence of mental health disorders and the growing adoption of digital interventions. Over recent years, the integration of game elements into mental health applications has evolved from early serious games to sophisticated gamified platforms leveraging smartphone capabilities and virtual reality, reflecting a trajectory from basic engagement tools to complex therapeutic modalities [1,2,3]. This evolution is significant given that mental health conditions affect a substantial portion of the global population, with up to 20% of children and adolescents experiencing disorders such as anxiety and depression [4,5]. Digital gamified interventions offer scalable, accessible, and potentially cost-effective solutions to address treatment gaps, especially in underserved populations [6,7].

Despite promising developments, specific challenges persist in understanding how gamification impacts user engagement, therapeutic efficacy, and ethical considerations within digital mental health contexts. While some studies report enhanced engagement and symptom reduction through gamified cognitive behavioural therapy and serious games [8,9], others highlight heterogeneity in outcomes and methodological inconsistencies [10,11]. Moreover, ethical concerns regarding privacy, autonomy, and potential adverse effects such as addiction remain underexplored, particularly in vulnerable groups like adolescents [12,13,14]. Competing perspectives debate the balance between the motivational benefits of gamification and its risks of over-engagement or misuse [15,16]. The lack of standardised frameworks and comprehensive evaluations limits the ability to generalise findings and optimise intervention design [17,18].

To synthesise these disparate findings, this review introduces and applies the “Engagement–Efficacy–Ethics Trilemma” as a novel conceptual framework. We posit that the design of gamified DMHIs involves inherent trade-offs between three competing vertices: (1) Engagement, the maximisation of user adherence and retention; (2) Efficacy, the achievement of meaningful clinical outcomes; and (3) Ethics, the upholding of principles of user autonomy, privacy, and beneficence. While established models like the Fogg Behaviour Model (FBM) or the Persuasive Design Category (PDC) focus on the mechanics of behavioural change, the Engagement–Efficacy–Ethics Trilemma extends these by explicitly integrating the clinical and ethical trade-offs often overlooked in technical HCI research. Unlike previous deductive models, this framework emerged inductively from a synthesis of current DMHI literature. The resulting conceptual model, along with the specific tensions between its vertices, is illustrated in Figure 1.

As illustrated in Figure 1, the trilemma is defined by three distinct mechanisms of tension that govern the design and impact of DMHIs:

(1)

The Engagement–Efficacy Tension: This represents the ‘Engagement–Efficacy Gap’ identified in the literature, where design optimised for momentary behavioural adherence results in high ‘app engagement’ (e.g., logins, time-on-app) but fails to trigger the deep cognitive processing required for ‘therapeutic engagement’ and symptom reduction.

(2)

The Engagement–Ethics Tension: This occurs when persuasive mechanics, such as variable reward schedules or competitive leaderboards, leverage cognitive biases to sustain use. These risks cross the line from supportive nudging into cognitive manipulation, potentially undermining user autonomy or inducing psychological distress.

(3)

The Efficacy–Ethics Tension: This tension is frequently amplified by intelligent systems, where AI-driven personalisation relies on invasive data collection and ‘black-box’ algorithms. While such tailoring can increase clinical efficacy, it often creates a conflict with ethical principles of transparency, privacy, and user agency.

This paper will systematically deconstruct this trilemma, using it as a lens to analyse the current evidence, explore the amplifying role of Artificial Intelligence (AI) and the mediating role of Human–Computer Interaction (HCI), and propose a path toward a more integrated and responsible approach to design and evaluation.

Specifically, this review aims to systematically synthesise empirical evidence and theoretical perspectives on gamified Digital Mental Health Interventions (DMHIs) through the lens of the Engagement–Efficacy–Ethics Trilemma, identifying key design tensions, methodological gaps, and future research priorities. The paper is structured as follows: Section 2 outlines the key theoretical foundations of gamification. Section 3 details the systematic review methodology. The main findings are then presented across four sections that deconstruct the trilemma: Section 4 examines the Engagement–Efficacy Gap, Section 5 explores the ethical landscape, Section 6 analyses the role of AI as an amplifier, and Section 7 discusses HCI as a mediator. Finally, Section 8 provides a general discussion of the findings, and Section 9 concludes with recommendations for responsible innovation in digital mental health.

2. Theoretical Foundations of Gamification

The motivational appeal of gamification is rooted in established psychological theories of human behaviour. Understanding these foundations is crucial for analysing their application in a therapeutic context.

2.1. Behaviourism and Operant Conditioning

Many common gamification elements, such as points, badges, and rewards, are direct applications of behaviourist principles, specifically Skinner’s theory of operant conditioning [19]. These elements function as positive reinforcement, delivered on a schedule to encourage the repetition of desired behaviours, such as completing a daily therapeutic exercise [20]. Immediate feedback and small rewards can trigger the release of dopamine, creating a powerful reinforcement loop that motivates continued use [21]. However, this approach primarily fosters extrinsic motivation, which may not be sustainable in the long term and can lead to user fatigue if not carefully balanced [20].

2.2. Self-Determination Theory (SDT)

A more nuanced perspective is offered by Self-Determination Theory (SDT), which posits that intrinsic motivation—the drive to engage in an activity for its inherent satisfaction—is fostered by satisfying three core psychological needs: Autonomy (a sense of control), Competence (a feeling of mastery), and Relatedness (a feeling of social connection) [22]. Effective gamification can be designed to support these needs. For example, customisable avatars can enhance autonomy [8], progress bars and levels can support competence [18], and social leaderboards can foster relatedness [23]. A design focus that supports these innate psychological needs is more likely to lead to sustained, internally motivated engagement than one based purely on extrinsic rewards [16].

2.3. The Fogg Behaviour Model (FBM)

The Fogg Behaviour Model (FBM) [24] provides a practical framework for designing persuasive technology by positing that a behaviour (B) occurs when three elements converge simultaneously: Motivation (M), Ability (A), and a Prompt (P). The model is often expressed as the formula B = MAP. FBM is particularly useful in the context of gamification as it provides a clear diagnostic and design tool for encouraging user actions and has been used to guide the selection of game elements in health interventions.

• Motivation: This aligns with the drives discussed in SDT and Behaviourism. Fogg categorises core motivators into three pairs: Sensation (pleasure/pain), Anticipation (hope/fear), and Belonging (social acceptance/rejection). Gamification directly targets these, for instance, by offering rewards (pleasure) or creating leaderboards (social belonging).
• Ability: Fogg reframes ability as simplicity. A behaviour is more likely to be performed if it is easy to do. Gamification can increase a user’s ability by breaking down complex therapeutic tasks into smaller, more manageable steps, such as quests or levels, thereby reducing the required physical or mental effort.
• Prompt: A behaviour will not happen without a trigger or cue. Gamified systems are rich with prompts, from push notifications that remind users to complete a daily challenge to in-app visual cues that prompt the following action.

The FBM is valuable because it shows that motivation and ability are trade-offs; a user with very high motivation might complete a difficult task, but a user with low motivation will only complete a very simple one. For designers of DMHIs, this provides an actionable insight: to ensure a therapeutic behaviour occurs, they can focus on making the action as simple as possible, which is often more practical than trying to increase a user’s motivation perpetually. Together, Behaviourism, Self-Determination Theory, and the Fogg Behaviour Model provide a complementary theoretical foundation for understanding the motivational mechanisms that underlie gamified Digital Mental Health Interventions. This synthesis highlights how the interplay between extrinsic reinforcement, intrinsic motivation, and behavioural triggers can inform both the design and ethical evaluation of these systems.

3. Methodology of Literature Selection

The review process followed the PRISMA framework (Figure 2) to ensure transparency and reproducibility in study identification, screening, and inclusion. The objective was to identify, synthesise, and critically appraise studies concerning the influence of gamification on engagement, efficacy, and ethics in digital mental health. The completed PRISMA 2020 Checklist, detailing the location of all reporting items within this manuscript, is provided in Supplementary Materials.

3.1. Search Strategy and Information Sources

A comprehensive literature search was conducted across multiple electronic databases, including MEDLINE (via PubMed), PsycINFO, Web of Science, and Scopus, for articles published between January 2015 and June 2025. The search strategy combined keywords and subject headings related to three core concepts: (1) Gamification, (2) Digital Mental Health, and (3) the outcomes of interest. An example search string used for Scopus was: (TITLE-ABS-KEY(gamif*) OR TITLE-ABS-KEY(“serious games”)) AND (TITLE-ABS-KEY(“mental health” OR “digital health” OR mHealth)) AND (TITLE-ABS-KEY(engag* OR efficac* OR ethic* OR adher*)).

3.2. Eligibility Criteria

Studies were included if they met the following criteria: (1) were empirical studies (e.g., randomised controlled trials, quasi-experimental studies), systematic reviews, or meta-analyses; (2) focused on interventions using gamification or game-based elements for mental health purposes; (3) reported on outcomes related to user engagement, clinical efficacy, or ethical considerations; and (4) were published in English. Exclusion criteria were strictly defined to eliminate ‘noise’ and ensure clinical validity. We excluded: (1) conference abstracts, dissertations, and grey literature to maintain strong evidence quality. Preprints were screened to capture emerging evidence, but they were not considered primary evidence in the synthesis. (2) pilot studies lacking statistical power analysis or reporting solely on user experience (UX) and usability without clinical outcomes; and (3) general ‘wellness’ apps not targeting specific psychiatric disorders (e.g., DSM-5 criteria) to ensure relevance to therapeutic efficacy. This rigorous filtering process resulted in a final selection of 50 high-quality empirical studies. To address the risk of bias, all 50 included studies were critically appraised using the Mixed Methods Appraisal Tool (MMAT (2018 Version)) to ensure the robustness of the synthesis, even where a formal meta-analysis of trial data was not performed. The MMAT appraisal was used to inform the critical interpretation of findings and assess methodological robustness, rather than as a strict exclusion threshold.

3.3. Study Selection and Data Extraction

Following the removal of duplicates, titles and abstracts were independently screened by two reviewers. Full texts of potentially relevant articles were then retrieved and assessed for final inclusion. A citation chaining process, encompassing both backwards (reviewing reference lists) and forward (identifying citing articles) searches, was conducted on key systematic reviews to identify additional relevant studies.

Data were extracted from the final included studies using a standardised form, capturing information on study design, population, intervention characteristics, gamification elements, and key findings related to engagement, efficacy, and ethics. Included studies were qualitatively coded based on three dimensions: (1) Engagement (quantified by retention rates or app usage frequency); (2) Efficacy (measured by standardised clinical symptom scales such as the PHQ-9 or GAD-7); and (3) Ethics (identified through reported discussions on privacy, autonomy, or adverse behavioural effects). This structured synthesis provided the inductive basis for identifying the tensions within the Trilemma. Any discrepancies between reviewers during the screening or data extraction stages were resolved through discussion and consensus.

The systematic search and selection process resulted in the inclusion of 50 studies. A descriptive summary of these studies is provided in Appendix A (

Table A1. Descriptive Summary of Key Studies.

Study	Engagement Metrics	Therapeutic Efficacy	Ethical Frameworks	Gamification Design Features	User Experience and Adherence
[1]	High engagement via smartphone serious games; frequent use reported	Positive psychological outcomes; varied symptom targets	Limited ethical discussion; focus on usability and engagement	Interactive, immersive, user-tailored game mechanics	Good adherence; self-administrable interventions
[10]	Mixed engagement data: reward and progress elements are common	Positive effects on well-being and depressive symptoms; heterogeneous for anxiety and stress	Ethics are briefly mentioned; need for rigorous designs	Rewards, sensations, and progress elements are prevalent	Variable retention; inconsistent engagement metrics
[5]	Increased engagement in paediatric populations compared to traditional therapy	Early evidence of therapeutic benefit for anxiety, depression, and ADHD	Ethical aspects are not deeply explored	Gamified video game-based interventions	Enhanced adherence relative to non-gamified treatments
[77]	Engagement is enhanced by goal setting, feedback, and social interaction	Improved patient outcomes across healthcare domains	Ethical concerns include autonomy, privacy, and addiction risks	Narrative immersion, progress tracking, and personalisation	Sustained participation linked to design features
[17]	Engagement is emphasised via the target audience and the user engagement focus.	Mechanisms of action linked to health effectiveness	Ethical tensions between healthcare and entertainment paradigms	Gamification, serious games, purpose-shifted entertainment games	User engagement is critical for intervention success
[17]	Focus on engagement through paradigm integration	Therapeutic effects linked to game mechanics	Ethical challenges in balancing entertainment and health goals	Customisation, gamification, serious game design	Engagement mediates efficacy; design impacts adherence
[78]	High completion rates: app used frequently in short sessions	Significant reductions in negative emotions and maladaptive coping	Ethical considerations include usability and feasibility	Gamified mobile app with metacognitive skill training	Positive user feedback: feasibility for RCTS confirmed
[6]	64.5% adherence; higher retention than controls	Significant improvements in resilience, anxiety, and depression	Ethics is not the primary focus; it is implied in design	Gamified mobile app teaching psychological skills	High user retention and satisfaction reported
[79]	Gamification increased app usage frequency and points earned	Positive correlation between engagement and mental health improvement	Ethical issues were not explicitly addressed	Points, badges, leaderboards, gamification	Engagement is positively influenced by gamification
[8]	Avatar customisation increased in-the-moment engagement	Customisation improved training efficacy, reduced anxiety	Ethical implications implicit in user autonomy and personalisation	Avatar customisation as motivational design	Increased identification enhanced adherence and efficacy
[80]	The gaming component encouraged goal adherence in serious mental illness	Feasibility study underway; expected health improvements	Ethical focus on patient motivation, adherence, and autonomy	Rewards based on individualised goals	Early indications of good adherence and engagement
[9]	A gamified CBT app increased usage and retention in child anxiety	Improved skill practice and engagement compared to non-gamified	Ethical considerations include acceptability and usefulness	Interactive games, rewards, and a messaging interface	Higher engagement and retention than prior versions
[2]	Engagement improved via motivational dynamics in applied games	Promising evidence for depression treatment efficacy	Ethics are discussed in terms of research gaps and user needs	Variety of game types, including exergames and cognitive training	User-centred design emphasised for adherence
[81]	Moderate engagement; PC-based serious games	Moderate effect size on symptom improvement	Limited ethical discussion; focus on clinical outcomes	Goal-oriented and cognitive training games	Feasibility and accessibility issues noted
[12]	Engagement is discussed in the context of young people’s use	Ethical aspects are central; limited efficacy data	Comprehensive ethical framework, including privacy and autonomy	Serious gaming elements with co-design emphasis	Ethical integration linked to sustained use
[13]	Engagement is linked to social connection	Ethical concerns about harms and gaps	In-depth ethical analysis, including privacy and justice	Ethical design principles are recommended	User trust and safety are emphasised for adherence
[14]	Engagement increased, but with risks of addiction and privacy invasion	Ethical concerns about behavioural manipulation	Strong focus on ethical frameworks and privacy	Gamification risks highlighted	Calls for responsible and ethical gamification
[15]	Motivation and commitment improved via gamification	Ethical and moral aspects are critical in design	Framework for ethical gamification development	Balance of motivation and user well-being	Ethical design supports sustained engagement
[27]	Gamification increased motivation and enjoyment	Supported efficacy for depression, anxiety, and stress	Ethics briefly noted; focus on effectiveness	Combination of gaming and CBT techniques	Positive adherence linked to gamification
[26]	Engagement metrics varied; persuasive design principles analysed	Significant clinical improvements in mental health outcomes	Ethics not the primary focus: design principles evaluated	Use of persuasive design elements	Engagement data is inconsistently reported
[23]	Frequent use of customisation, narrative, and feedback elements	Positive outcomes in anxiety, depression, and stress reduction	Ethical considerations implicit in design choices	Extended reality and game-based intervention elements	User motivation supported by immersive features
[82]	High engagement in school-based gamified health promotion	Effective in reducing anxiety, depression, and burnout	Ethical issues addressed via CBT and neuropsychological integration	CBT and neuropsychological principles are embedded	Positive user adherence and behavioural outcomes
[74]	High retention (>90%) and adherence (>80%) in Ugandan adolescents	Trends toward symptom reduction; feasibility confirmed	Ethical acceptability emphasised via co-production	Narrative gamification with telephonic guidance	High acceptability and sustained use reported
[7]	Large-scale reach with moderate completion rates	Knowledge gains and positive user feedback	Ethical considerations implicit in accessibility	Interactive voice response game format	Strong acceptability: engagement improvements needed
[83]	Engagement is comparable between gamified and group CBT	Both interventions are effective for anxiety and depression	Ethics not primary focus: clinical effectiveness emphasised	Gamified self-guided CBT app and group training	Good adherence and symptom remission reported
[84]	The feasibility study showed good recruitment and retention	Small to large effect sizes on mental health measures	Ethical acceptability supported by qualitative feedback	ACT-based video game with therapeutic focus	Positive user experience and adherence
[18]	User feedback guided gamification design for engagement	Gamification enhanced intrinsic motivation and behavioural change	Ethical design integrated with behavioural theories	Personalisation, challenges, and assistance elements	High engagement and user satisfaction reported
[85]	VR gamified CBT increased engagement in university students	Reduced short-term anxiety; usability confirmed	Ethical considerations implicit in design	VR with CBT techniques and gamification	Positive user experience and usability
[86]	Gamified CBT app tailored for Arabic users	Reduction in depression and anxiety symptoms	Cultural and ethical tailoring emphasised	Gamified CBT elements adapted culturally	User satisfaction and symptom improvement
[87]	Gamification integrated with CBT in a narrative game	Supports depressive symptom reduction via engagement	Ethical review by CBT experts included	Narrative, verbal, physical, and social media gamification	Positive user engagement and learning
[20]	Gamification improved retention via rewards and levels	Increased intrinsic motivation and regular use	Ethical design stressed avoiding over-reliance on extrinsic motivators	Points, badges, and leaderboards used	User-centred design critical for sustained use
[75]	Cultural and clinical factors influence engagement	Feasibility and acceptance in Malaysian adults with depression	Ethical concerns about addiction and stigma are addressed	Storylines, therapist guidance, safety measures	Acceptance is linked to culturally sensitive design
[16]	Engagement varies with personalisation and gender factors	Effectiveness is influenced by study design and client needs	Ethical risks of over-engagement and addiction are discussed	Personalisation vs. standardisation debated	Tailored gamification is recommended for adherence
[88]	Increased gamification use during COVID-19: engagement rising	Supports mental health improvement during the pandemic	Ethics are briefly mentioned; there is a need for further research	Mobile and web-based gamified platforms	Engagement facilitated by remote access
[89]	A gamified educational game improved mental health literacy	Reduced stigma and increased awareness	Ethical focus on education and stigma reduction	Educational content integrated with gamification	Positive user feedback and community impact

). At the same time, a critical appraisal of their methodological quality is detailed in Appendix B. Thematic patterns across these studies, which informed the development of the Trilemma, are synthesised in Appendix C.

4. The Engagement–Efficacy Gap: The First Tension of the Trilemma

The primary justification for implementing gamification in DMHIs is its purported ability to solve the engagement crisis. A compelling body of evidence suggests that gamification is successful in achieving this primary aim, leveraging motivational design to sustain user engagement and participation. However, a critical analysis of the literature reveals a significant “Engagement–Efficacy Gap”, where increased app usage and interaction do not consistently or reliably translate into superior clinical outcomes. This paradox forms the first major tension of the trilemma.

A substantial portion of the reviewed literature (35 out of 50 studies) reports that gamification significantly enhances user engagement, motivation, and retention. Interventions that incorporate elements such as rewards, progress tracking, and customisation consistently demonstrate higher adherence and interaction time [1,20]. For example, a large-scale randomised controlled trial of the gamified app eQuoo found it retained 42% more participants than control groups, with an overall adherence rate of 64.5% [6]. Similarly, features like avatar customisation have been shown to increase in-the-moment engagement and foster a stronger user connection to the intervention [8].

While the evidence for increased engagement is strong, the data on whether this translates into superior clinical efficacy is far more ambiguous. Meta-analyses and systematic reviews repeatedly find no significant difference in effectiveness for reducing symptoms of depression or anxiety between gamified and non-gamified apps [10]. For instance, a meta-analysis of mental health app RCTs found that while interventions effectively reduced depressive symptoms overall (Hedges g = −0.27), the number of gamification elements was not a significant predictor of efficacy ($\beta$ = −0.03, $\rho$ = 0.38) [25]. These quantitative values are cited to illustrate consistent patterns reported in prior meta-analyses, rather than to constitute a new quantitative synthesis conducted within this review. This demonstrates a clear ‘Engagement–Efficacy Gap’ numerically: the presence of game mechanics does not statistically improve outcomes compared to non-gamified versions. Furthermore, meta-regression indicates that gamification elements do not significantly impact intervention adherence ($\beta$ = −1.93, $\rho$ = 0.40), directly challenging the assumption that ‘pointsification’ alone solves the engagement crisis [25]. The divergence between engagement metrics and clinical outcomes is summarised in Table 1, which juxtaposes evidence of the Engagement–Efficacy Gap across a subset of high-impact studies.

Some reviews even report no significant effect of gamification on intervention adherence, directly challenging the simpler narrative [26]. The literature is characterised by a high degree of heterogeneity in outcomes for conditions like anxiety and stress, with many studies suffering from small sample sizes and short follow-up periods, which limits the robustness of efficacy claims [11,27].

This discrepancy suggests a critical distinction between “app engagement” and “therapeutic engagement”. The former, often measured by logins or time spent on the app, can be driven by extrinsic motivators that may only lead to superficial interaction with game mechanics. The latter requires deep cognitive and emotional processing of therapeutic content. The risk is that users may learn to “game the system” for rewards without engaging in the more effortful work required for clinical improvement, thus creating a gap between high activity metrics and stagnant therapeutic progress. This divergence underscores the need for a multidimensional definition of engagement that distinguishes between behavioural, cognitive, and affective components, each contributing differently to therapeutic outcomes.

Table 1. Juxtaposing Evidence for the Engagement–Efficacy Gap.

Study	Key Findings on Engagement	Key Findings on Efficacy
[6]	High adherence (64.5%) and 42% higher retention in the gamified app group compared to the control group.	Significant improvements in resilience, anxiety, and depression in the gamified group.
[8]	Avatar customisation increased in-the-moment engagement and user identification with the intervention.	Customisation improved training efficacy and was associated with reduced anxiety.
[9]	A gamified Cognitive Behavioural Therapy (CBT) app for child anxiety led to increased usage and higher retention compared to a non-gamified version.	The gamified version resulted in improved therapeutic skill practice.
[25]	A meta-regression showed no significant effect of gamification on intervention adherence.	A meta-analysis found no significant difference in effectiveness for reducing depressive symptoms between gamified and non-gamified apps.
[10]	Engagement data was mixed; reward and progress elements were common, but metrics were inconsistent.	Positive effects on well-being and depressive symptoms, but heterogeneous and inconsistent results for anxiety and stress.
[28]	A systematic review found no statistically significant difference in adherence between different gamification features or the number of features.	The review did not conduct a meta-analysis of clinical efficacy but noted that usage data were not commonly reported, which limited the conclusions that could be drawn.

Table 1. Juxtaposing Evidence for the Engagement–Efficacy Gap.

Study	Key Findings on Engagement	Key Findings on Efficacy
[6]	High adherence (64.5%) and 42% higher retention in the gamified app group compared to the control group.	Significant improvements in resilience, anxiety, and depression in the gamified group.
[8]	Avatar customisation increased in-the-moment engagement and user identification with the intervention.	Customisation improved training efficacy and was associated with reduced anxiety.
[9]	A gamified Cognitive Behavioural Therapy (CBT) app for child anxiety led to increased usage and higher retention compared to a non-gamified version.	The gamified version resulted in improved therapeutic skill practice.
[25]	A meta-regression showed no significant effect of gamification on intervention adherence.	A meta-analysis found no significant difference in effectiveness for reducing depressive symptoms between gamified and non-gamified apps.
[10]	Engagement data was mixed; reward and progress elements were common, but metrics were inconsistent.	Positive effects on well-being and depressive symptoms, but heterogeneous and inconsistent results for anxiety and stress.
[28]	A systematic review found no statistically significant difference in adherence between different gamification features or the number of features.	The review did not conduct a meta-analysis of clinical efficacy but noted that usage data were not commonly reported, which limited the conclusions that could be drawn.

Overall, the evidence suggests a persistent Engagement–Efficacy Gap that challenges assumptions about the therapeutic value of gamification, setting the stage for examining the ethical tensions that arise when persuasive design succeeds in sustaining use without demonstrable clinical benefits.

5. The Ethical Landscape: The Second Tension of the Trilemma

The pursuit of engagement, particularly when it does not guarantee efficacy, forces a confrontation with the third vertex of the trilemma: ethics. Gamified DMHIs are persuasive technologies designed to influence the behaviour of a vulnerable population, a context that demands rigorous ethical scrutiny. The literature reveals that the very mechanics used to drive engagement are often the source of the most significant ethical risks, establishing a direct tension between maximising user retention and the ethical imperative to “not harm”.

A significant portion of the reviewed literature (22 out of 50 studies) raises critical ethical concerns. Yet, these discussions are often underdeveloped, focusing narrowly on research ethics while neglecting broader socio-political implications [12,13]. The most frequently cited ethical concern is the potential for psychological distress, which can manifest as stress from unattainable goals, anxiety from constant feedback, or demotivation from social comparison on leaderboards [25,26,27,29,30]. These findings illustrate that ethical risks are not peripheral side effects, but structural consequences of design choices aimed at sustaining engagement.

Furthermore, while persuasive design—grounded in ‘nudge theory’—can encourage positive behavioural change while respecting user autonomy, it exists on a spectrum. When design becomes opaque, it risks transitioning into cognitive manipulation. However, it is essential to acknowledge that for patients with severe depression, short-term extrinsic motivation (e.g., rewards) may serve as a `necessary short-term trade-off’ to encourage the first steps toward engagement, provided long-term autonomy is not compromised [14]. This is particularly problematic when the underlying persuasive mechanisms are opaque to the user, as illustrated in

Table 2. Mapping Engagement Mechanics to Documented Ethical Harms.

Category of Harm	Specific Harm	Causal Gamification Mechanic/Principle	Supporting Literature
Psychological	Anxiety and Stress	Unattainable goals, high-frequency feedback, and constant performance pressure from features like leaderboards.	[25,26,27,29,30]
Autonomy	Cognitive Manipulation	Leveraging cognitive biases (e.g., loss aversion, scarcity) and opaque algorithms to influence user behaviour without their full awareness.	[14,30,31]
Privacy	Data Exploitation/Surveillance	Continuous tracking of sensitive personal and health data is required for personalisation and feedback loops.	[12]
Behavioural	Addiction/Compulsive Use	Use of variable reward schedules and compelling feedback loops designed to maximise time-on-app rather than therapeutic benefit.	[15]
Social	Negative Social Comparison	Public leaderboards and competitive features that rank users against one another can potentially exacerbate feelings of inadequacy.	[13]

. The continuous data tracking required for personalisation and feedback loops poses significant privacy risks and can evoke feelings of surveillance, while also increasing the likelihood of data exploitation and misuse [13,14,30,31]. Finally, the same design patterns that make apps “sticky” can foster addictive or compulsive behaviours, which are antithetical to the therapeutic goal of fostering self-control and well-being [15].

In sum, the ethical tension within the trilemma emerges from the paradox that the very mechanisms enhancing engagement may simultaneously compromise user autonomy, privacy, and psychological well-being. Addressing this tension requires an explicit ethical framework integrated into the design and evaluation of gamified DMHIs, rather than treating ethics as an afterthought.

6. AI as an Amplifier of the Trilemma

The emergence of Artificial Intelligence (AI), particularly the recent integration of Large Language Models (LLMs), is reshaping the landscape of gamified DMHIs. AI amplifies the trilemma through specific mechanisms: for example, ‘black box’ algorithmic personalisation can increase efficacy through extreme tailoring while simultaneously reducing ethics by masking the logic behind interventions. This lack of Explainable AI (XAI) exacerbates the tension between therapeutic success and user transparency [32,33,34,35,36,37,38,39]. AI is often positioned as the key to resolving the shortcomings of static, one-size-fits-all interventions by introducing a new level of personalisation and adaptability that could potentially close the Engagement–Efficacy gap [4,32,33,34,35,36,37,38,39]. However, this same capability simultaneously intensifies the ethical tensions of the trilemma to an unprecedented degree [40,41]. This dual amplification effect positions AI not merely as a tool within the trilemma but as an active agent that reshapes its boundaries and dynamics.

AI-driven systems promise to enhance efficacy by crafting highly personalised treatment plans in real-time [41]. By analysing vast datasets, including user behaviour, self-reported moods, and even biometric data from wearable devices, AI algorithms can dynamically adjust the difficulty of therapeutic challenges, recommend specific interventions when signs of distress are detected, and tailor feedback to an individual’s personality [4,36,37,38,39,42,43,44]. This capacity for deep personalisation holds the promise of making treatments more engaging and, crucially, more clinically relevant, and effective [32,33,34,35]. Conversational agents, also known as chatbots, represent a prominent application, offering 24/7 accessibility and a non-judgmental space for users to practice coping skills. Platforms like Woebot and Wysa have demonstrated effectiveness in reducing symptoms of depression and anxiety [45,46,47].

This pursuit of personalisation, however, magnifies the ethical risks exponentially. AI-driven adaptation requires even more invasive and continuous data collection, deepening privacy concerns and creating what could be termed a “hyper-personalised panopticon” [48,49,50,51,52,53,54]. The algorithms themselves are often opaque “black boxes”, making it difficult for users or clinicians to understand why a particular recommendation was made, which raises profound issues of accountability and algorithmic bias [41,48,49,50,51,53,54]. Such biases, often stemming from unrepresentative training data, can lead to AI systems that systematically underdiagnose or misclassify individuals from marginalised groups, thereby perpetuating existing health disparities [55,56]. The potential for manipulation also becomes far more sophisticated; an adaptive system can learn an individual’s psychological vulnerabilities and tailor its persuasive strategies accordingly, making manipulation more effective and almost impossible to detect [48,49,50,51,52,53,54]. This creates a fundamental power asymmetry between the user and a system that not only tracks behaviour but also actively evolves its strategies to capture and hold their attention [38,41,57,58,59].

Table 3. The Dual Role of AI in the Gamification Trilemma.

AI Function	Example Application	Potential Benefit (to Efficacy)	Potential Risk (Ethical Concern)
Personalisation/Adaptation	Dynamically adjusting the complexity of CBT exercises based on user performance; recommending meditations based on mood logs.	Increases therapeutic relevance and prevents user boredom or frustration, potentially closing the Engagement–Efficacy gap [32,33,34,35].	Requires invasive data collection; risk of algorithmic bias; potential for hyper-personalised manipulation [41,48,49,50,51,52,53,54].
Conversational Agents (Chatbots)	Woebot or Wysa provides CBT through guided conversations and offers 24/7 crisis support.	Increases accessibility of care, reduces stigma, and provides continuous support [47].	Lack of genuine empathy; risk of providing inappropriate advice in a crisis; potential for emotional dependency [48,49,50,51,52,53,54].
Player and Behavioural Analytics	Identifying which game mechanics lead to the highest user retention and promoting them within the app.	Can improve the overall user experience and increase long-term engagement [32,33,34,35].	The optimisation goal (engagement) may conflict with the therapeutic goal, leading to the creation of addictive or compulsive loops.
Early Detection and Screening	Analysing speech patterns for signs of depression or decision-making patterns in a game to screen for Attention-Deficit/Hyperactivity Disorder (ADHD).	Enables early detection and intervention; provides objective data to supplement clinical assessment [41,42,43,44].	Risk of misdiagnosis and stigmatisation due to algorithmic bias; significant data privacy concerns [55,56].

illustrates this multifaceted role of AI in the gamification trilemma.

Explainable AI (XAI) as a Potential Mitigator

In response to the ethical challenges posed by opaque “black box” models, the field of Explainable AI (XAI) has emerged as a critical area of research [60,61,62,63,64,65]. XAI refers to methods and techniques that make the outputs of machine learning models understandable to humans, providing clear and interpretable explanations for how and why specific decisions were made. In the context of mental health, XAI can promote trust, transparency, and accountability [60,61,62,63]. For clinicians and patients to trust an AI’s recommendation, they need to understand its reasoning, moving beyond a simple prediction to a comprehensible rationale [64,65,66].

However, implementing XAI is not without its own challenges. There is often a trade-off between a model’s performance and its interpretability; the most accurate deep learning models are often the most complex and challenging to explain. Furthermore, the explanations themselves must be meaningful to the end-user (e.g., a clinician or patient), not just to a data scientist [64,65,66]. While XAI offers a promising path to mitigating some of the trilemma’s ethical risks by making AI systems more transparent and accountable, it is not a complete solution. It remains an active area of research [60,63,65,66]. Future research should therefore explore how principles of Explainable and Responsible AI can be systematically embedded into the design of gamified DMHIs, ensuring that personalisation enhances rather than undermines ethical integrity.

7. HCI as the Mediator of the Trilemma

The abstract tensions of the Engagement–Efficacy–Ethics trilemma are not decided in theory but are negotiated in the tangible reality of the user’s interaction with the digital interface. Human–Computer Interaction (HCI) is the discipline that governs this reality, and its principles and practices are where design choices can either exacerbate the trilemma’s conflicts or mediate them to create more balanced, effective, and ethical interventions [67,68].

A central challenge in digital mental health is fostering a “digital therapeutic alliance” (DTA)—the sense of trust, collaboration, and connection that is a key predictor of positive outcomes in traditional therapy [69]. HCI design choices, from the tone of a chatbot’s language to the perceived responsiveness of the system, play a critical role in building or hindering this alliance by fostering empathy and trust [69]. The literature underscores that a lack of interpersonal factors is a primary reason for disengagement, highlighting the need for HCI to focus not just on technical usability but on designing for relational connection [67]. This shift reframes HCI from a focus on usability metrics to one of ethical mediation, where every design element carries both therapeutic and moral weight.

To better analyse how these systems function, it is useful to deconstruct the user’s experience through an HCI lens that categorises different modes of interactivity. These modes reveal how every design decision is an implicit negotiation of the trilemma:

• Human-to-Data Interaction: The user inputs data (e.g., a mood log) and receives a visualisation back. This can enhance competence but raises privacy concerns about the data being logged.
• Human-to-Human Interaction: The system mediates connections between people, such as in peer support forums or competitive leaderboards. While this can foster relatedness, leaderboards are a clear example of a design choice that prioritises engagement at the potential cost of ethical well-being (by inducing anxiety) and efficacy (by distracting from therapeutic goals).
• Human-to-AI Interaction: The user interacts with an intelligent agent, such as a therapeutic chatbot. The design of this interaction directly mediates the tension between personalised efficacy and the ethical risks of manipulation and a lack of genuine empathy.
• Human-to-Algorithm Interaction: The user interacts with the underlying game mechanics, such as levelling up or earning points. The design of these algorithms determines whether the motivation is primarily extrinsic (potentially undermining long-term efficacy) or intrinsic (supporting autonomy and competence).

Table 4. HCI Design Choices as a Negotiation of the Trilemma.

HCI Interaction Mode	Example Design Choice	Potential Positive Mediation (Supporting Efficacy/Ethics)	Potential Negative Consequence (Prioritising Engagement)
Human-to-Data	Mood tracking with reflective journaling prompts vs. simple emoji logging.	Encourages deeper therapeutic engagement and self-awareness.	Simplistic logging is faster and easier, boosting “app engagement” but offering little therapeutic depth.
Human-to-Human	Moderated peer-support forum.	Fosters a sense of relatedness and community, supporting long-term well-being.	It can be less immediately “engaging” than a competitive feature.
Human-to-Human	Competitive Leaderboard.	It can motivate some users through social competition.	Induces anxiety, negative social comparison, and distracts from personal therapeutic goals.
Human-to-AI	Empathetic chatbot with transparent, explainable logic (XAI).	Builds trust and a strong digital therapeutic alliance, enhancing efficacy.	A purely engagement-optimised chatbot might use manipulative language or hide its logic to keep the user talking.
Human-to-Algorithm	Skill-based progression system that unlocks content based on mastery.	Supports competence and ensures the user is ready for the following therapeutic step.	A time-based or point-based system is easier to implement and can drive daily logins, but may not align with therapeutic progress.

illustrates how specific HCI design choices within different interaction modes represent a direct negotiation of the trilemma’s tensions.

Ultimately, HCI is the discipline of applied ethics for these complex systems. The principles of user-centred design and co-design, which involve users with lived experience and clinicians throughout the development process, are presented in the literature as the most promising path to mitigating the harms of the trilemma [12,18]. By prioritising the user’s therapeutic needs and ethical safety over simplistic engagement metrics, HCI can guide the development of interventions that are not just sticky but genuinely supportive. In this sense, HCI serves as the operational domain where the abstract principles of engagement, efficacy, and ethics are translated into concrete design decisions. Strengthening this translational bridge through participatory and theory-driven design represents a crucial step toward resolving, rather than merely managing, the trilemma [70,71,72,73].

8. Discussion

This systematic review has synthesised evidence across clinical psychology, information ethics, AI, and HCI to construct and apply the Engagement–Efficacy–Ethics Trilemma as a critical framework. The complexities identified in this review often stem from conflicting results across diverse populations; a detailed Analysis of Agreement and Divergence (AAD) in the existing literature is presented in Appendix E. The analysis reveals that the current paradigm, often characterised by a siloed focus on maximising engagement, is fundamentally flawed. Specifically, our findings substantiate the three core mechanisms of tension established in Section 1. The “Engagement–Efficacy Tension” is evidenced by the consistent disconnect between adherence metrics and symptom reduction. Furthermore, the “Engagement–Ethics Tension” and “Efficacy–Ethics Tension” are clearly manifested in the way persuasive design and AI-driven black-box personalisation often compromise user autonomy and transparency in favour of “stickiness” and tailored outcomes. This integrated framework moves beyond descriptive mapping toward a normative model that defines what responsible and effective gamification should look like in digital mental health.

The findings call for a paradigm shift away from designing for simple “app engagement” and toward designing for “therapeutic engagement”. This requires moving beyond extrinsic motivators and prioritising designs that foster intrinsic motivation by supporting users’ needs for autonomy, competence, and relatedness. Furthermore, the field must shift from a reactive to a proactive ethical stance, embedding ethical risk assessment into every stage of the design process. A comprehensive roadmap for addressing these unresolved tensions is detailed in Appendix G, which categorises gaps and future research directions by priority level.

Based on the literature, several key research gaps must be addressed to resolve the tensions of the trilemma. There is a pressing need for the standardisation of engagement metrics [10], the execution of more longitudinal studies to assess long-term effects [10,20], and the development of comprehensive ethical frameworks that go beyond basic research ethics [10,12,13,14]. Future research should focus on isolating the specific mechanisms by which different game elements impact therapeutic outcomes and prioritise the cultural adaptation of interventions through participatory co-design with diverse user groups [18,74,75].

A detailed thematic analysis and chronological evolution of these research directions are summarised in Appendix C and Appendix D. Specifically, our data extraction identifies a significant trend toward ‘pointsification’ with 70% (35/50) of included studies leveraging extrinsic motivators such as rewards and progress tracking to drive participation (Appendix C). However, the chronological mapping in Appendix D reveals that while earlier works (2012–2015) focused primarily on feasibility and initial engagement, the most recent literature (2024–2025) illustrates an increasing reliance on immersive technologies and AI-driven personalisation, which often intensifies the trilemma’s ethical tensions by prioritising tailored efficacy over user transparency.

9. Conclusions

The accumulated literature on gamification in digital mental health interventions reveals a compelling consensus that gamification enhances user engagement, boosts therapeutic efficacy, and offers novel opportunities for mental health promotion and treatment. However, this review demonstrates that its application is far more complex than initially conceived. The Engagement–Efficacy–Ethics Trilemma provides a crucial conceptual lens for navigating this complex territory, revealing the inherent trade-offs between keeping users engaged, achieving clinical outcomes, and upholding ethical principles. The accelerating integration of Artificial Intelligence promises to push this trilemma to its limits, offering unprecedented opportunities for personalised efficacy alongside unprecedented risks of manipulation and surveillance. Ultimately, the future of gamified DMHIs depends on a fundamental reorientation toward responsible innovation. By embracing principles of co-design, prioritising ethics, fostering intrinsic motivation, and demanding transparency, it may be possible to harness the immense potential of these technologies. To do so, we must design not only for the user’s attention, but also for their autonomy, well-being, and dignity. Resolving the Engagement–Efficacy–Ethics Trilemma requires a systemic convergence of technological innovation, psychological theory, and ethical design—a triadic balance that should define the next generation of digital mental health research and practice. Our synthesis of the current dataset identifies that addressing these tensions requires a shift from feature-centric to motivation-centric design (Appendix H). The broader theoretical and practical implications arising from this synthesis, including the need for standardised engagement metrics and proactive ethical risk assessment, are expanded upon in Appendix F, providing a foundational roadmap for sustainable innovation in this domain.