Computational and Mathematical Methods in Medicine

Computational and Mathematical Methods in Medicine / 2020 / Article

Research Article | Open Access

Volume 2020 |Article ID 3573267 |

R. Loula, L. H. A. Monteiro, "A Game Theory-Based Model for Predicting Depression due to Frustration in Competitive Environments", Computational and Mathematical Methods in Medicine, vol. 2020, Article ID 3573267, 6 pages, 2020.

A Game Theory-Based Model for Predicting Depression due to Frustration in Competitive Environments

Academic Editor: Hans A. Braun
Received10 Jan 2020
Revised22 Apr 2020
Accepted08 May 2020
Published03 Jun 2020


A computational model based on game theory is here proposed to forecast the prevalence of depression caused by frustration in a competitive environment. This model comprises a spatially structured game, in which the individuals are socially connected. This game, which is equivalent to the well-known prisoner’s dilemma, represents the payoffs that can be received by the individuals in the labor market. These individuals may or may not have invested in a formal academic education. It is assumed that an individual becomes depressed when the difference between the average payoff earned by the neighbors in this game and the personal payoff surpasses a critical number, which can be distinct for men and women. Thus, the transition to depression depends on two thresholds, whose values are tuned for the model accurately predicting the percentage of individuals that become depressed due to a frustrating payoff. Here, this tuning is performed by using data of young adults living in the United Kingdom in 2014-2016.

1. Introduction

Major depressive disorder, the mental illness popularly called depression, affects over 300 million people worldwide [1]. One relevant risk factor for depression is the lack of access to education [2]. Around 2010, in the United Kingdom, the access to higher education was facilitated [3]. However, in the following years, instead of decreasing, depression prevalence increased among men and women in the age range of graduating and starting to work [35]. Surveys conducted in 2014 [4] and 2016 [5] revealed that about 20% of young adults suffered from depression. Here, a computational model is proposed to investigate this issue.

The influence of job attributes (such as working conditions, workload, wage, stability, and career opportunities) on the occurrence of depression has been examined [6, 7]. Here, a negative perception of the career progress is considered a risk factor for depression. Thus, depression is supposed to be a consequence of unfulfilled expectations in the professional career. In the model, the insertion of young adults in their socioeconomic environment is represented by a game, which is shown to be equivalent to the prisoner’s dilemma [8, 9].

Game theory has been employed in psychological studies, for instance, to analyze the influence of a depressive mood on economic decisions modeled by the ultimatum game [10] and to describe psychiatric disorders as Nash disequilibria of the self, reflecting discordances between actions and beliefs [11]. The game introduced here relates investment in formal education to professional success. It is supposed that, in a competitive environment, the individual becomes depressed when the difference between the average earning received by the neighbors in this game and the personal earning is above a threshold. Since this critical number can be different for men and women, then there are two thresholds ruling the transition to depression. The values of these parameters can be adjusted from real-world statistics, in order to improve the prediction accuracy of the model.

An individual-based probabilistic model was already proposed to forecast the prevalence of depression from a set of negative life events affecting a population, by taking into consideration its demographic and socioeconomic features [12]. Other approaches for detecting and predicting depression employ, for instance, artificial intelligence techniques [1315] and signal processing tools [1618]. Here, the model is formulated in terms of a game with spatial structure, a state transition rule (from healthy to sick), and a parameter adaptation algorithm.

This paper is organized as follows. In Section 2, the proposed model is introduced. In Section 3, the results obtained in computational simulations based on the data of young adults living in the United Kingdom in 2014-2016 are presented. In Section 4, the possible relevance of these results is discussed.

2. The Model

Consider a population of young adults (that is, individuals in the 18–24-year-old age range). Let this population be represented by a square lattice , in which the top and bottom edges are connected and the right and left edges are also connected. This boundary condition is chosen to avoid edge effects. Each cell composing this toroidal lattice corresponds to an individual, who maintains social contact with the eight surrounding neighbors. In the literature on cellular automata, this regular coupling pattern is known as the Moore neighborhood of unit radius [19]. This regular graph has been employed, for instance, in works about the spreading of contagious diseases [20, 21]. The individuals of the proposed model have gender, in order to take into account that the number of women affected by depression is higher [22, 23]. Such individuals are randomly distributed over the lattice according to the gender statistics.

Suppose that each individual had played once the investing-in-education game “against the field.” In a game against the field, there is no specific opponent. Here, the field is supposed to be the labor market. The investing-in-education game is summarized in Figure 1. The strategies that can be chosen by the individuals are E (academically educated, that is, student) and NE (nonacademically educated, that is, nonstudent). The strategy E is adopted by those who invest money and time in a standard academic education. The strategy NE is employed by those who do not care or do not have access to a formal education. The strategies provided by the field are S (success) and NS (nonsuccess). The strategy S means professional success (reaching a high socioeconomic status); NS means professional failure.

As shown in Figure 1, each individual receives one of the following payoffs: for (NE,S), for (E,S), for (NE,NS), and for (E,NS). Indeed, these payoffs are randomly distributed by the population, according to official statistics when available. For instance, if 10% of individuals have not had a formal education and did not achieve professional success, then the payoff is randomly assigned to 10% of the population, regardless of the individual’s gender.

It is reasonable to assume that success without investment in education is better than that with investment; therefore, . Unsuccess without investment is better than that with investment; hence, . Also, success with investment is better than unsuccess without investment; thus, . In short, . Note that these inequalities concerning the payoffs (with ) can also be found in the famous prisoner’s dilemma [8, 9]. In this classic game theory example, two suspects are caught and accused of jointly committing a crime. When they are separately interrogated, their two options are D (defection) or C (cooperation). The strategy D means to confess to committing the crime; C means to remain silent. The payoff corresponds to the temptation to defect against a cooperating partner, is the reward for mutual cooperation, is the punishment for mutual defection, and is the sucker’s payoff, which is the outcome of cooperating when the partner defects [8, 9]. The equivalence between the games can be established by replacing the individual and the field in the investing-in-education game by the two suspects involved in the prisoner’s dilemma and by replacing the strategies E and S by C and the strategies NE and NS by D. Because of this equivalence, a rational individual who believes in destiny should not study (in this context, destiny is the lack of causal connection between formal education and success), since NE is the dominant strategy (giving the higher payoff) when the field plays either S or NS. However, people usually consider that investment in education is correlated with economic prosperity [24]. Notice that the mentioned equivalence is not perfect, since, in practice, the field does not (directly) receive any payoff; therefore, the investing-in-education game is not symmetric, as the prisoner’s dilemma is.

To add variability, there are two kinds of individuals, who are denoted by A and B, in the proportions and , respectively. The payoffs earned by them are different when they are professionally successful. It is assumed that and . Therefore, the parameters of this spatial game are the values of , , , , , and and the percentages of individuals who have gotten these payoffs.

Let be the payoff earned by the -th individual and the average earning of the neighborhood (which is calculated from the values of of the eight neighbors). All individuals are initially supposed to be healthy. The -th individual becomes sick when: (1) for men(2) for women

Observe that progression to depression occurs when the difference between the neighborhood payoff and the personal payoff is above a threshold, which can be different for men and women. Thus, the professional competence is subjectively assessed by comparing the wage in the current job to the neighbors’ wages. A negative perception can lead to depression.

The thresholds and are tuned in the calibration phase. Assume that the variable expresses the time steps of this phase. Let be the depression prevalence determined from the model at and , the official estimate. From the initial values and , is computed by initially taking and . In the model, is the maximum absolute error allowed in the prediction of . Usually, it is necessary to adjust these two thresholds, in order to obtain such an error. The parameter adaptation algorithm is given by: (1)if , then and (2)if , then and in which is a positive constant. When the official number is achieved (with maximum error equal to ), the ratio between the prevalence in women and in men is calculated. This ratio must be between 1.5 and 3.0 [4, 22, 23]. If it is outside this interval, then is subtracted from the current value of and a new step of the calibration phase is executed. The calibration finishes at the time step , for which and the ratio of sick women to sick men is between 1.5 and 3.0. Then, the threshold values are taken as and , and numerical simulations are performed in order to obtain , the average value of in these simulations.

In short, the simulation dynamics is the following. A gender and a payoff are randomly assigned (according to official statistics when available) to each individual of the lattice. He/she compares the average payoff of the neighborhood to his/her personal payoff. An individual becomes depressed if the difference between these payoffs is above a threshold. Since this threshold can be different for men and women, then, in fact, there are two thresholds, which are denoted by and . These parameters are tuned in the calibration phase. The parameter adjustment ends when the target percentage of depressed individuals and the expected ratio between the prevalence in women and in men are achieved. The values of and found in the calibration phase are assumed to be psychological characteristics of the studied population. Hence, after the calibration, the model can be used to determine, for instance, how the average prevalence (obtained in simulations) depends on the fractions of individuals receiving the six payoffs.

3. Numerical Results

Numerical simulations were performed with the payoffs , , , , , and . Also, (therefore, the population is composed of 10000 individuals), (thus, there are 60% of A individuals and 40% of B individuals), (the maximum absolute error), (the parameter used to tune and in the calibration phase), (the number of simulations performed after finishing the calibration), and and (the initial values of the parameters ruling the transition to depression).

The target of the simulations is , which is considered to be the prevalence of depression (due to personal frustration) in young adults in the United Kingdom in 2014-2016 [4, 5]. The gender distribution was 49.3% male and 50.7% female [25]. The percentage of nonstudents was 18.8% [26] (recall that nonstudents can receive the payoffs , , or ). Consequently, the percentage of students was (and students can receive , , or ). Assume that is the percentage of successful nonstudents (who received or ) and is the percentage of successful students (who received or ). With these choices, the prevalence predicted by the model was , which is close to the target [4, 5]. Also, and , and the ratio between the prevalence in women and in men given by the model was 1.95, which belongs to the interval .

In this first set of simulations, the percentages and of successful individuals (who studied and who did not study, respectively) were arbitrarily chosen. To evaluate the impact of these percentages in the depression prevalence , other sets of simulations were run. These results are exhibited in Figures 25.

Figure 2 presents an expected outcome: decreases by increasing (and consequently by reducing in order to keep the percentage of nonstudents constant. Recall that ). Thus, depression prevalence decreases by increasing the fraction of individuals receiving or (and concomitantly by reducing the fraction of individuals receiving ).

The plots shown in Figure 3 are surprising: the relation between and is not monotonous. The dashed line corresponds to , which can represent a socioeconomic environment. In the labor market, a small parcel of individuals does achieve professional success without investing in an academic formation (for instance, a digital influencer with a natural talent for humor or a very skilled soccer player). The solid line corresponds to , which can represent a learning environment. In a university, there is no academic achievement without studying. In both plots, the depression prevalence increases up to and, only above this number, decreases (note that, as increases, decreases, because remains constant). Therefore, depression prevalence increases and then decreases by increasing the fraction of individuals receiving or (and concomitantly by reducing the fraction of individuals receiving ).

Figure 4 shows that the higher the , the higher the . Thus, the higher the fraction of nonstudents, the higher the prevalence of depression.

The parameter values in Figures 24 are the same as those used in the first set of simulations. To get a full picture of this issue, additional sets of simulations were performed by varying the percentages , , and . In Figure 5, the darker the circle, the higher the . Observe that, for other values of and (different from those used in Figure 3), has its maximum for intermediate values of .

4. Discussion and Conclusions

Depression due to social competition was already examined from an evolutionary point of view, through a discussion of the function of its mental state over the course of the human evolution [27]. Here, the occurrence of depression in competitive individuals was investigated from an educational-socioeconomic standpoint. It is undeniable that formal education promotes intellectual development, enhancing the opportunities in the labor market. However, many newly graduated students face unemployment and low occupational status; in other words, they have difficulties in finding a good job [3, 28, 29]. In fact, a higher education diploma favors, but does not guarantee, social progress. Also, analyzing the benefit-cost ratio concerning this investment can become a stressful experience, capable of triggering mental health problems, such as anxiety, hostility, and depression [3, 28, 29].

In the United Kingdom, depression prevalence increased in young adults, after the access to higher education being stimulated [35]. This observation can be explained by the proposed model, in which psychological distress emerges from comparisons of payoffs. Thus, a negative assessment of the personal success can lead to depression.

Figure 4 shows the expected relationship between the depression prevalence and the percentage of nonstudents ; that is, increases with . The relationship between and the percentage of successful nonstudents shown in Figure 2 was also not surprising; that is, decreases with . However, the dependence between and the percentage of successful students presented in Figure 3 is not obvious. Figure 5 reveals that this behavior can also be found for other parameter values.

As shown in Figure 3, starts decreasing with only above a critical number (about 70% in this figure). Therefore, from a government perspective, the access to higher education should be even more stimulated, in order to overcome this critical number. From an individual perspective, analysis of the benefit-cost ratio on investing in education should be avoided early in the career, because such a premature analysis can lead to wrong conclusions. In fact, the usual lack of emotional maturity of young adults can increase their vulnerability to stressful thoughts, like concerns about their jobs. Therefore, risk of depression can be reduced by setting realistic aspirations and by avoiding competitive comparisons of professional success.

Data Availability

The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest

The authors declare that there are no conflicts of interest regarding the publication of this paper.


RL thanks the Instituto Presbiteriano Mackenzie for the scholarship. LHAM is partially supported by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) under the grant #304081/2018-3. This study was financed in part by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) (finance code 001).


  1. World Health Organization, Depression and Other Common Mental Disorders: Global Health Estimates, WHO, Geneva, Switzerland, 2017.
  2. J. Licinio and M. L. Wong, Eds., Biology of Depression: From Novel Insights to Therapeutic Strategies, Wiley, Weinheim, Germany, 2005. View at: Publisher Site
  3. S. Cant, “Hysteresis, social congestion and debt: towards a sociology of mental health disorders in undergraduates,” Social Theory & Health, vol. 16, no. 4, pp. 311–325, 2018. View at: Publisher Site | Google Scholar
  4. S. McManus, P. Bebbington, and R. Jenkins, Mental Health and Wellbeing in England: Adult Psychiatric Morbidity Survey, T. Brugha, Ed., vol. 2016, NHS Digital, Leeds, UK, 2014.
  5. M. Smith, “One in four students suffer from mental health problems,” YouGov, 2016, View at: Google Scholar
  6. R. Steyn and N. Vawda, “Job characteristics: their relationship to job satisfaction, stress and depression,” Journal of Psychology in Africa, vol. 24, no. 3, pp. 281–284, 2014. View at: Publisher Site | Google Scholar
  7. R. Axelrod, The Evolution of Cooperation, Basic Books, New York, USA, 2006.
  8. G. Szabó and G. Fáth, “Evolutionary games on graphs,” Physics Reports, vol. 446, no. 4-6, pp. 97–216, 2007. View at: Publisher Site | Google Scholar
  9. K. M. Harlé and A. G. Sanfey, “Incidental sadness biases social economic decisions in the ultimatum game,” Emotion, vol. 7, no. 4, pp. 876–881, 2007. View at: Publisher Site | Google Scholar
  10. T. Patokos, “The relevance of Nash equilibrium to psychiatric disorders,” Theoretical Medicine and Bioethics, vol. 32, no. 4, pp. 245–258, 2011. View at: Publisher Site | Google Scholar
  11. R. Loula and L. H. A. Monteiro, “An individual-based model for predicting the prevalence of depression,” Ecological Complexity, vol. 38, pp. 168–172, 2019. View at: Publisher Site | Google Scholar
  12. W. Mumtaz, S. S. A. Ali, M. A. M. Yasin, and A. S. Malik, “A machine learning framework involving EEG-based functional connectivity to diagnose major depressive disorder (MDD),” Medical & Biological Engineering & Computing, vol. 56, no. 2, pp. 233–246, 2018. View at: Publisher Site | Google Scholar
  13. X. Li, R. La, Y. Wang et al., “EEG-based mild depression recognition using convolutional neural network,” Medical & Biological Engineering & Computing, vol. 57, no. 6, pp. 1341–1352, 2019. View at: Publisher Site | Google Scholar
  14. S. A. Akar, S. Kara, S. Agambayev, and V. Bilgiç, “Nonlinear analysis of EEGs of patients with major depression during different emotional states,” Computers in Biology and Medicine, vol. 67, pp. 49–60, 2015. View at: Publisher Site | Google Scholar
  15. H. Jiang, B. Hu, Z. Liu et al., “Detecting depression using an ensemble logistic regression model based on multiple speech features,” Computational and Mathematical Methods in Medicine, vol. 2018, Article ID 6508319, 9 pages, 2018. View at: Publisher Site | Google Scholar
  16. M. Bachmann, L. Päeske, K. Kalev et al., “Methods for classifying depression in single channel EEG using linear and nonlinear signal analysis,” Computer Methods and Programs in Biomedicine, vol. 155, pp. 11–17, 2018. View at: Publisher Site | Google Scholar
  17. S. Wolfram, Cellular Automata and Complexity: Collected Papers, Westview Press, Boulder, USA, 1994.
  18. J. C. A. Dias and L. H. A. Monteiro, “Clustered Breeding Sites: Shelters for Vector-Borne Diseases,” Computational and Mathematical Methods in Medicine, vol. 2018, Article ID 2575017, 5 pages, 2018. View at: Publisher Site | Google Scholar
  19. D. F. Ferraz and L. H. A. Monteiro, “The impact of imported cases on the persistence of contagious diseases,” Ecological Complexity, vol. 40, Article ID 100788, 4 pages, 2019. View at: Publisher Site | Google Scholar
  20. American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, American Psychiatric Association, Washington, USA, 2013. View at: Publisher Site
  21. S. Marginson, “The worldwide trend to high participation higher education: dynamics of social stratification in inclusive systems,” Higher Education, vol. 72, no. 4, pp. 413–434, 2016. View at: Publisher Site | Google Scholar
  22. Office for National Statistics, Population by age, gender and ethnicity, ONS, 2017,
  23. Offce for National Statistics, “Measuring national well-being in the UK: international comparisons,” ONS, p. 2019, 2019, View at: Google Scholar
  24. P. B. Jackson and M. Finney, “Negative life events and psychological distress among young adults,” Social Psychology Quarterly, vol. 65, no. 2, pp. 186–201, 2002. View at: Publisher Site | Google Scholar
  25. M. Grenfell, Pierre Bourdieu: Key Concepts, Routledge, New York, USA, 2014. View at: Publisher Site
  26. F. J. Zimmerman, D. A. Christakis, and A. Vander Stoep, “Tinker, tailor, soldier, patient: work attributes and depression disparities among young adults,” Social Science & Medicine, vol. 58, no. 10, pp. 1889–1901, 2004. View at: Publisher Site | Google Scholar
  27. Y. Lee, R. M. Ragguett, R. B. Mansur et al., “Applications of machine learning algorithms to predict therapeutic outcomes in depression: a metaanalysis and systematic review,” Journal of Affective Disorders, vol. 241, pp. 519–532, 2018. View at: Publisher Site | Google Scholar
  28. M. Piccinelli and F. Gomez Homen, Gender Differences in the Epidemiology of Affective Disorders and Schizophrenia, WHO, Geneva, Switzerland, 1997.
  29. J. Price, L. Sloman, R. Gardner Jr., P. Gilbert, and P. Rohde, “The social competition hypothesis of depression,” British Journal of Psychiatry, vol. 164, no. 3, pp. 309–315, 1994. View at: Publisher Site | Google Scholar

Copyright © 2020 R. Loula and L. H. A. Monteiro. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

More related articles

 PDF Download Citation Citation
 Download other formatsMore
 Order printed copiesOrder

Related articles

Article of the Year Award: Outstanding research contributions of 2020, as selected by our Chief Editors. Read the winning articles.