Problematic internet usage: can commitment and progress frameworks help regulate daily personal internet use?

ABSTRACT

Objective

Problem Internet Usage (PIU) is associated with numerous psychological concerns. The dynamics of self-regulation (DSR) model may provide a useful framework for psychological interventions with PIU, given previous research showing that the framework can be successfully applied to PIU behaviours.

Method

The authors conducted a randomised controlled trial (RCT) to evaluate the efficacy of an internet-based intervention to reduce daily personal internet hours, and PIU severity, for university students who reported PIU issues. Assessments were undertaken at baseline, on each day of the 21-day intervention, and 6-week follow-up. Seventy-four participants completed the intervention and 38 completed follow-up assessments.

Results

The experimental group reduced daily personal internet usage to a greater degree than the active control group at the end of the intervention. Findings also demonstrated a greater reduction in Internet Addiction Test (IAT) scores for the experimental group compared to the active control, with a third of participants in the experimental condition demonstrating clinically significant change in IAT scores.

Conclusions

Our findings suggest that the DSR framework may provide a promising approach to reducing PIU.

KEY POINTS

What is already known about this topic:

1. Problematic internet use has been viewed in various ways, including as an addiction or a problem of self-regulation.

2. Interventions to reduce problematic internet use, including medication and cognitive-behavioural therapy, have shown limited success.

3. The Dynamics of Self-Regulation Model has been shown to influence intended internet behaviours.

What this topic adds:

1. The Dynamics of Self-Regulation Model was applied to daily internet use in a randomised controlled trial.

2. Daily personal internet usage reduced to a greater degree using an experimental intervention drawn from the Dynamics of Self-Regulation Model compared to an active control intervention of self-monitoring.

3. Internet addiction scores also reduced using the experimental intervention drawn from the Dynamics of Self-Regulation Model compared to self-monitoring.

KEYWORDS:

• Problem internet usage
• self-regulation
• internet addiction, intervention, translational research

The increase in internet and smartphone usage over the past decade (American Psychological Association, 2017; International Telecommunication Union, 2019; Internet World Stats, 2019) has led to the modern issue of Problematic Internet Usage (PIU). Numerous studies have displayed direct links between excessive internet use and unfavourable outcomes for individuals across a range of psychological of areas of psychological functioning, including depression, loneliness, obsessive compulsive disorder (OCD), attention-deficit hyperactivity disorder (ADHD), generalised anxiety disorder, social anxiety, sleep disorders, suicide ideation, academic achievement and relationship quality (Cheng et al., 2018; Muusses et al., 2014; Spada, 2014; Wang et al., 2014; Weinstein et al., 2015; Yau et al., 2013).

PIU has been viewed as a pathology or addiction (Young, 1996, 1998), a cognitive-behavioural problem (Davis, 2001), a socio-cognitive construct (LaRose & Eastin, 2004; LaRose et al., 2003), or as self-regulation failure in controlling mood and behaviours (Tokunaga, 2016). Interventions to reduce PIU have shown limited success, including medication (usually prescribed for depression or ADHD), cognitive behavioural therapy (individual or group), or family-based therapy (Cash et al., 2012; King et al., 2011; Winkler et al., 2013).

A model of the dynamics of self-regulation (DSR) developed by Ayelet Fishbach and colleagues over the last decade (Fishbach & Dhar, 2005; Fishbach & Zhang, 2008, 2008; Fishbach et al., 2006, 2009; Zhang et al., 2007) has the potential to inform interventions for reducing PIU. The theories have mostly been tested and employed in marketing and consumer studies to date (Campbell & Warren, 2015; Fitzsimons et al., 2008; Wilcox et al., 2009) but recent research (Dunbar et al., 2017, 2018) showed that essential components of the model can be employed in the PIU domain.

The Dynamics of Self-Regulation (DSR) model includes a dual representation framework of goal-directed behaviour with simultaneous pursuit of multiple goals and temptations and their effects on resultant behavioural outcomes. Fundamental to the model is the proposition that individuals assess their progress in moving towards, or level of commitment to, a focal goal when regulating multiple goals (Fishbach & Dhar, 2005). A commitment representational framework motivates individuals to track and regulate their commitment level to the goal end state. Positive behaviour outcomes suggest a strong commitment to the focal goal and the subsequent dynamic of self-regulation is highlighting congruent future goal behaviours. In contrast, an unsuccessful goal behaviour denotes a lack of commitment to the goal, leading to doubting the importance of the goal, and raising the level of appeal for temptations or incongruent goal behaviours. For example, asking people how committed they feel to reducing internet use and linking reduced internet use to a productive life would highlight and increase interest in behaviours congruent with that goal. Unsuccessful regulation of internet use would indicate a lack of commitment and may lead to doubts about the importance of the goal. A progress goal representational framework motivates individuals to monitor and regulate the difference between the present and preferred end state. According to this framework, successful goal behaviours indicate partial achievement of the goal and signify that enough exertion towards goal completion has been achieved. The resulting dynamic of self-regulation is that of balancing and the individual feels free to temporarily disengage from the focal goal, resulting in the increased salience of other goals and behaviours. As an example, asking people if their successful regulation of internet use was helping them make progress towards their goal of decreasing internet use would decrease their interest in subsequent regulation of internet use. Unsuccessful regulation of internet use, conversely, indicates a discrepancy between the present and preferred end state and motivates subsequent regulation of internet use. Asking people in this situation how much progress they feel they have made after recent unsuccessful performance would motivate subsequent regulation of internet use.

Current study

The DSR model has been applied to PIU, showing that intended behaviours for individuals could be primed and influenced (Dunbar et al., 2017, 2018). Dunbar et al. (2017) showed that when temptations (internet behaviours) and goal congruent actions (academic behaviours) are presented together and appear to complement each other, university students value temptations more highly than when the same actions are presented apart and appear to compete against each other. For example, studying for a quiz (academic) was linked with browsing shopping websites (internet behaviours) by an “and then” statement. On the other hand, when temptations and goal congruent actions are presented apart and appear to compete against each other (e.g., studying for a quiz “OR” browsing shopping sites), university students value goal congruent actions more highly than when the same actions are presented together and appear to complement each other.

In three separate studies, Dunbar et al. (2018) showed that success or failure feedback about productivity, inducing high or low engagement by providing information about PIU or no information, and presenting abstract goals to reduce personal internet usage, or concrete goals to reduce internet usage by a certain amount, could induce either a commitment or progress framework in university students.

We set out to expand on those studies by applying the theory to behaviour in a real world setting using a randomised controlled trial design, and extended the previous research by moving from a general population to a population of university students screened for PIU. We aimed to evaluate the effectiveness of an online intervention for university students with problematic internet usage. Participants who were engaged with the goal of reducing internet usage were recruited. The online intervention included DSR elements of success feedback, commitment framing and presentation of internet behaviours as competing, if personal internet use was reduced, and failure to progress feedback and progress framing if personal internet use was not reduced. It was hypothesised that the experimental intervention would be more effective in reducing personal internet hours compared with an active control group. Recent reviews of self-monitoring have found it is an effective behaviour change technique across a variety of domains (Harkin et al., 2016; Rose et al., 2017). In order to balance the environments and expectations for both groups, we used an active control group that employed self-monitoring. We also posited that participants in the experimental group would show a greater change in scores on the Internet Addiction Test compared to those in the active control group. A six-week follow-up was conducted to explore any lasting effects of the intervention over time.

The following hypotheses were generated from previous research and the dynamics of self-regulation model (Dunbar et al., 2017, 2018; Fishbach et al., 2009):

H1.

Participants in the experimental group will have a greater reduction in daily personal internet (DPI) hours than those in the active control group at the end of day 21.

H2.

Participants in the experimental group will reduce their Internet Addiction Test (IAT) score significantly more than those in the active control group at the end of Day 21.

H3.

Participants in the experimental group will have a greater reduction in daily personal internet (DPI) hours than those in the active control group at six-week follow-up.

H4.

Participants in the experimental group will reduce their Internet Addiction Test (IAT) score significantly more than those in the active control group at six-week follow-up.

Materials and methods

Design

The study was a parallel group, 21-day randomised controlled trial (RCT) with a six-week follow-up. All participants who met the inclusion criteria were allocated to intervention or active control using block allocation (Kang et al., 2008) so that the group sizes never differed by more than two participants. Participants were blinded to group allocation, but the experimenter was not. The outcome measures were changes from baseline in self-reported daily personal internet (DPI) hours and Internet Addiction Test (IAT) scores. The study was conducted in accordance with the CONSORT standards (Eysenbach & Group, 2011; Moher et al., 2012), and the flow of participant progress through the phases of the study is shown in Figure 1.

Figure 1. Flow of participants (CONSORT flow chart). IAT: internet addiction test.

Participants

Undergraduate and postgraduate university students (N = 247; 53% female, 47% male, M_age = 25.01 years, SD = 7.88) were recruited via the university web portal with an advertisement asking for individuals interested in decreasing their amount of personal internet usage. Individuals who clicked on the link were taken to the screening survey, where they were given information relating to the study, were asked for consent, entered demographic information, completed the screening test for PIU and gave an estimate of their daily personal internet hours. The target population is individuals who experience difficulties regulating their time on the internet. Therefore, participants who scored 40 or higher on the IAT were included (Kuss & Lopez-Fernandez, 2016).

Participants were asked if they were interested in undertaking a 21-day study and informed that withdrawal from the study was possible at any time without consequence. After screening, 94 participants were randomly allocated to the experimental group (54% female, 46% male, M_age = 25.30 years, SD = 6.90) or active control group (57% female, 43% male, M_age = 23.90 years, SD = 7.40).

Procedure

The intervention relies on priming mental representation frameworks via the common understandings and social constructs of the English words commitment and progress and previous research (Dunbar et al., 2018) has found that this does not occur reliably with participants for whom English is a second language. Therefore, students for whom English is not their first language were excluded.

Participants were offered an incentive of being able to win one of four AUD100 gift vouchers if they completed the 21 days in full. Participants received an email each day at 7 am and were asked to report their daily personal internet (DPI) hours from the previous day. Participants were explicitly instructed to exclude time spent on the internet for academic or work purposes.

Participants in the experimental group were given theory-driven (Dunbar et al., 2017, 2018; Fishbach & Zhang, 2009; Fishbach et al., 2009) feedback depending on the outcome of the comparison of their last two days personal internet hours. If there was a reduction in daily personal internet hours it was considered a success, and commitment feedback was provided. Examples of commitment feedback include “Congratulations. Your commitment to reduce your internet usage is evident”, “After the success of the previous day, how committed do you feel to your goal to reduce your personal internet usage?” and “Consider when faced with a dilemma whether to use the internet for personal use, you can use the internet for personal use OR do something much more productive”. If the value was the same or worse than previously it was considered a failure, and progress feedback was provided. Examples of progress feedback include “Your results indicate you failed to progress towards your goal to reduce your personal internet usage. This indicates that improvement is required” and “How much progress towards your goal do you feel you have made after the disappointment of yesterday?”. In order to prevent repetition, four variations of each feedback scenario of commitment and progress conditions were created and randomly selected for each participant on each occasion.

Participants in the active control group were provided with self-monitoring feedback which included showing them their hours for the last 2 days. Again, four versions were created and randomly presented to each participant. Examples of self-monitoring feedback include “Yesterday you spent XXX hours on personal internet usage while the day before you spent XXX hours on the internet for personal use” and “After the results of yesterday, how much motivation towards pursuing your goal do you feel?” The texts presented to the experimental and active control groups were approximately equal in length.

On the final day (day 21), the IAT was also administered. Participants were thanked for their time and informed that they would be contacted in 6 weeks’ time for a final follow-up. At follow-up, daily personal internet hours and the IAT scales were collected.

Measures

Internet Addiction Test (IAT; Young, 2015) is a widely accepted and validated tool and provides useful cut-off scores distinguishing individuals with problematic internet usage from those without (Kuss & Lopez-Fernandez, 2016; Pontes et al., 2016; Škařupová et al., 2015). The IAT is a self-report 20-item scale using a 5-point Likert scale ranging from 1 (“Rarely”) to 5 (“Always”). Scores of 20–39 indicate normal internet use, whereas scores of 40–69 indicate frequent problems and scores of 70–100 indicate significant problems with internet usage (Kuss & Lopez-Fernandez, 2016). The IAT has excellent internal consistency with Cronbach’s alpha typically above 0.90 across studies, has good to excellent concurrent validity with many other PIU scales and has convergent validity with time spent online (Laconi et al., 2014).

Statistical analysis

A linear mixed effects model was constructed to assess the effect of the intervention on daily personal internet hours. This model gives more statistical power than other techniques such as ANCOVA, is equipped to accommodate missing values at the various time-points at which the data was collected, and results in minimal loss of information as every PIU value recorded for every participant contributes to the analysis (Egbewale et al., 2014; Magezi, 2015; Meteyard & Davies, 2020; O’Connell et al., 2017). Analysis of covariance (ANCOVA) was used to assess the effect of the intervention on Internet Addiction Test (IAT) scores, controlling for pre-intervention IAT score, and adjusting for age and gender.

All statistical analyses were conducted using Stata. The level of statistical significance was set at 0.05. Based on previous research (e.g., Dunbar et al., 2017, 2018; Fishbach & Dhar, 2005), a priori power analysis was carried out before data collection to determine the required sample sizes. The required sample sizes were computed using the G Power computer program (Faul et al., 2007) with moderate effect size of 0.6, α of 0.05 and power of 0.80. Results suggested an estimated 90 participants in total or 45 per group would be required.

Ethics

The study procedures were carried out in accordance with the Declaration of Helsinki. The Human Research Ethics Committee of the University granted ethical approval (H-2018-016). Participants who expressed interest in reducing their daily internet usage were recruited and needed to meet criteria for having issues regulating their personal internet time. Further to this, every participant was informed that the study concerned reduction of personal internet usage hours. All participants gave written consent.

Results

Participant characteristics and outcomes by intervention group are summarised in Table 1. No differences were found between experimental and active control groups on baseline IAT scores, t(72) = 1.20, p = 0.23, 95% CI [−6.99, 1.73], Cohen’s d = 0.27, daily personal internet (DPI) hours, t(72) = 0.48, p = 0.63, 95% CI [−0.70, 1.14], Cohen’s d = 0.11, or Day-1 DPI hours, t(57) = 0.47, p = 0.64, 95% CI [−0.73, 1.17], Cohen’s d = 0.12.

Table 1. Participant characteristics and outcomes by intervention group.

	Experimental Group			Self-Monitoring Group
	N	M	SD	N	M	SD
Age in years	39	25.30	6.90	35	23.90	7.40
Gender (n, %)
Male	18 (46.15)			15 (42.86)
Female	21 (53.85)			20 (57.14)
Baseline 1 IAT score	39	51.08	8.15	35	53.71	10.60
Baseline 1 DPI hours	39	5.85	1.80	35	5.63	2.16
Day 1 DPI hours	33	4.20	1.88	26	3.98	1.70
Day 21 DPI hours	24	2.40	1.50	24	3.40	2.10
Follow up DPI hours	18	5.61	2.31	20	7.13	2.32
Day 21 IAT score	36	43.47	8.84	29	48.55	9.09
Follow up IAT score	17	42.82	9.98	20	51.70	13.04

IAT: Internet Addiction Test, DPI hours: daily personal internet hours.

On each of the 21 days of the intervention, participants were asked to report the number of DPI hours for the previous day. The mean reported DPI hours, at each time-point, for each group, are shown in Figure 2. The number of DPI values recorded per participant ranged from 0 (n = 1 participant) to 21 (complete DPI data, n = 17 participants).

Figure 2. Mean reported daily personal internet (DPI) hours at each day by intervention group.

The main fixed factor was intervention condition, and time was added as an interaction effect. Age and gender were set as fixed effects in the model. To account for repeated measures, participants were modelled as a random effect.

The estimated change in DPI hours over the 21 days showed that, for the experimental group, PIU decreased by 0.063 hours per day, p < .0001, whereas for the active control group, PIU decreased by 0.017 hours per day, p = 0.133. In order to determine if there is a difference between the slopes, the interaction term, time by group, was examined. The coefficient for the interaction term was −0.046, p = 0.003, meaning that DPI hours decreased by an extra 0.046 hours per day in the experimental group compared to the active control group, 95% CI [0.015, 0.077]. This supports Hypothesis 1.

The size of the difference in DPI hours between groups at any particular time point was also examined. As predicted, the difference between groups increased over time and the gap widened as the intervention continued. For example, on Day 2, the difference in groups is almost negligible, and DPI hours are 0.012 hours higher in the experimental group relative to the active control group, a difference of approximately about 42 seconds, p = 0.98. As the intervention continued, though, the gap widens and by Day 21, DPI hours were on average 52 minutes (0.87 hours) lower in the experimental group relative to the control group, p = 0.026, indicating there is evidence against the null hypothesis of no difference between groups. This also shows support for Hypothesis 1.

IAT data were measured at baseline (pre-intervention) and Day 21 (post-intervention), as shown in Table 1. There was a collective significant effect between baseline IAT, intervention condition, gender and age, F(4, 60) = 12.76, p < .0001, R² = 0.46. As expected, baseline IAT score, t(64) = 6.14, p < 0.001, 95% CI [0.41, 0.80] was a significant predictor in the model. Intervention group was also a significant predictor, t(64) = −2.39, p = 0.02, 95% CI [−7.71, −0.68]. On average, the intervention group scores for IAT were 4.20 units lower than that of the Active Control Group. These results support for hypothesis 2. Table 1 shows that pre- and post-intervention IAT scores for both experimental and active control groups demonstrated a reduction in IAT scores across the intervention. Scores on the IAT at Day-21 indicated that 19 of the participants, 13 (33.3%) from the experimental group and 6 (17.1%) from the active control group, had scores in the normal internet use range.

Six-week follow-up

The final sample completing the follow-up measures comprised n = 38 participants. Follow-up results are also shown in Table 1. At 6-week follow-up, there was a medium effect size that was not statistically significant between experimental and active control groups on follow-up daily personal internet (DPI) hours, t(36) = −1.96, p = 0.06, 95% CI [−3.08, 0.06], Cohen’s d = 0.66. However, differences were found between experimental and active control groups on follow-up IAT scores, t(35) = −2.23, p = 0.03, 95% CI [−16–96, −0.80], Cohen’s d = 0.76.

Given significant attrition from both groups at follow-up, a linear mixed-effects model was used to assess the effect of the intervention on Internet Addiction Test (IAT) scores, controlling for baseline IAT or DPI, age and gender. Intervention condition (experimental or control), time (in days, analysed as a continuous measure) and a condition-by-time interaction were entered as fixed effects, in addition to the adjustment variables. A random effect (random intercept) for participant was specified to account for repeated measurements from the same participant. Estimates of adjusted mean post-intervention IAT scores and DPI for each group were obtained from the model post-hoc, and differences in scores between groups were investigated at 6-week follow-up. All other factors were held fixed at their mean values.

At 6-week follow-up, the estimated mean DPI hours in the model for the experimental group were 3.60, 95% CI [2.94, 4.26], and for the active control group, it was 5.29, 95% CI [4.62, 5.96], signifying a difference between the groups of −1.69, 95% CI [−2.63, −0.75], p < 0.001, Cohen’s d = 0.56, supporting hypothesis 3. The estimated mean IAT score in the model for the experimental group was 43.27, 95% CI [39.17, 47.37] and for the active control group was 50.27, 95% CI [46.46, 54.08], a difference between the groups of 7.00, 95% CI [−12.6, −1.4], p = 0.015, Cohen’s d = 0.57. This offers support for hypothesis 4.

Discussion

The primary purpose of the study was to investigate the effect of the experimental intervention on reducing participants’ daily personal internet hours and their score on the Internet Addiction Test (IAT). Both groups reduced their daily personal internet hours, but the experimental group achieved this at a greater rate than that of the self-monitoring group, and this result was statistically significant with a medium effect size. Similarly, each group reduced their scores on the IAT, but by day-21 the experimental group had increased the gap to a statistically significant degree, with medium effect size. The hypotheses for both primary outcomes were supported, suggesting that PIU is indeed a self-regulation issue (LaRose et al., 2003; Spada, 2014; Yau et al., 2013). Our results support previous research in the area using the Dynamics of Self-Regulation model showing that it can be used to influence behaviours after successful and failed goal behaviours (Dunbar et al., 2017, 2018; Fishbach et al., 2009).

The data at 6-week follow-up showed that the experimental group maintained and extended the intervention effect on IAT scores, while the active control group scores trended towards baseline scores. Intervention effects showed a pattern of reversal towards baseline for DPI hours in both groups, although a significant difference was found between DPI hours for experimental and active control groups, indicating that the experimental group was still receiving benefit from the intervention. It is likely that participants who took the time to report at 6-week follow-up were also the most motivated and engaged participants and may not be a representative sample of the population. In addition, the sample size at follow-up was small, so the results should be interpreted with caution. Nevertheless, the results are encouraging and suggest that the intervention is worthy of further attention.

Aspects of the dynamics of self-regulation model, including presentation format, framing cues, feedback cues and focusing on the abstract goal versus concrete plans were combined to form the feedback provided to participants in the experimental group, and this intervention showed encouraging results. Given the apparent cost-effectiveness of similar interventions (Murray et al., 2016) and the potential for successful clinical outcomes, we recommend the dynamics of self-regulation be further investigated for its application to PIU. As a limitation, we note that the mean IAT score for the experimental group remained above the cut-off score of 40 at the end of the intervention and at follow-up, so that clinical significance was not demonstrated. Therefore, while the intervention showed promising results, future research should be based on a modified intervention, perhaps of longer duration, and augmented by increased experimenter contact as is discussed below.

With respect to the strengths of the study, there were several. Firstly, it was run in accordance with the CONSORT standards (Eysenbach & Group, 2011; Moher et al., 2012). Secondly, the interventions were carried out on a population of individuals experiencing problematic internet usage. Research in clinical populations has been lacking and considered a weakness in the PIU area (Tokunaga, 2017). Finally, an active control group condition was created to match the experimental group in order to balance attention, beliefs and expectations as much as possible between groups in or to prevent possible confounding issues (Simons et al., 2016).

With respect to the limitations of the study, an a priori power analysis determined that we would need 90 participants in order to find an effect. Despite 94 participants matching the inclusion criteria and being recruited, only 74 ultimately entered the study. Missing data were a factor throughout the study, although the rates of completion in the intervention phase (79%) are in line with similar internet-based interventions (Chebli et al., 2016). Our statistical approach, employing a linear mixed effects model, was utilised with this in mind as it gives more statistical power than other techniques, is better equipped to accommodate missing values, and results in minimal loss of information as every data point contributes to the analysis (Egbewale et al., 2014; Magezi, 2015; Meteyard & Davies, 2020; O’Connell et al., 2017). Nonetheless, it must be acknowledged that a small sample size limits the ability to draw conclusions regarding the effectiveness of the intervention.

While a longer term follow-up was desirable, resources did not allow for follow-up beyond 6 weeks. The mixed-effect model allowed for differences in the groups to be detected at 6-week follow-up but the sample size was small, and the results should be interpreted with caution. In addition, only 50% of participants who started the study completed follow-up measures, so that these participants may not be representative of the population. Therefore, the 6-week follow-up analysis should be considered exploratory rather than evidentiary. Although there was a degree of experimenter contact during the intervention phase regarding reporting of internet usage, there was no contact during the follow-up period. This may have contributed to the degree of dropout during the follow-up phase, as personal contact, even via email, was found to be useful in a review of internet-based interventions for similar problems (Chebli et al., 2016).

Future intervention studies using the DSR framework should include greater experimenter contact, particularly during the follow-up phase. This contact, which may be via email, could provide encouragement and should be separate from requests for reporting of internet use. Additional contact through telephone calls should also be considered, as it is associated with increased utilisation of intervention and greater effectiveness (Chebli et al., 2016). Should further trialling of the intervention show evidence of robust effects, it could potentially be implemented through student health settings or more widely as a readily available intervention on the internet. The degree of experimenter involvement required is a limitation on wider implementation, which could be addressed by increased automation of this role. However, decreased experimenter contact may potentially result in decreased utilisation and lower effectiveness. Therefore, wider implementation may potentially come with the cost of lower effectiveness.

With regard to methodological limitations, the study identified PIU based on a cut-off score indicating frequent problems with internet use, on the IAT, which measures excessive use, preoccupation and neglect of work among other attributes. It may not reflect other conceptualisations of PIU, about which there is as yet no consensus (Spada, 2014). The IAT is a self-report questionnaire, and this has the limitation of relying on the honesty and insight of participants. Future studies could include an interview assessment of PIU in order to obtain a more comprehensive assessment of PIU. In addition, the study relied on self-reported daily internet hours as a primary outcome, which may be subject to reporting biases. In addition, the study used retrospective data and was fixed in its timing of feedback to participants. However, participants were free to complete the questionnaire at any time of the day. Therefore, participants may have received feedback after a time when it could have influenced their behaviour. Future research could incorporate technology to deliver real-time monitoring and instant feedback.

Previous research showed that commitment or progress frameworks were not primed for participants for whom English was not the first language, as the effect depends on common understandings and social constructs of the English words commitment and progress (Dunbar et al., 2018, Study 1). Therefore, only individuals with English as their first language were included in this study. Given the extent of PIU across the globe (Tokunaga & Rains, 2016), this is a limitation of the present study, and future research could usefully examine if the current findings can be replicated in languages other than English. In addition, participants were drawn from a university population. This population is perhaps likely to use the internet to a greater extent than the general population, but it is unclear whether PIU is more or less prevalent than in a general population or a clinical population experiencing various mental disorders.

Conclusion

As almost two-thirds of the world’s population is now accessing the internet (Internet World Stats, 2019), and median ownership of smartphones in advanced countries is 76% and 45% in emerging countries (Silver, 2019), PIU is likely to be an increasing problem numerically, and across advanced and emerging countries. The person-based approach trialled in this study, based on the Dynamic Self-Regulation Model, is one possible approach that could be employed to address PIU. However, further investigation of a modified intervention based on the model is needed in order to obtain evidence of stronger effects of the intervention.

Ethical statement

This research was approved by the University of Adelaide Human Research Ethics School of Psychology Subcommittee, H-2018-016. All participants gave written consent. The study was conducted in accordance with the Declaration of Helsinki and according to requirements of the University of Adelaide Human Research Ethics School of Psychology Subcommittee.

Author contributions

DD led the study concept and design, conducted data collection, participated in data analyses and interpretation, wrote the first version of the manuscript, and edited the final manuscript. MP supervised study concept and design, supervised data interpretation, and wrote the final version of the manuscript. RR supervised study concept and design, supervised data interpretation, and edited the final version of the manuscript. The authors affirm that they had access to all data from the study, both what is reported and what is unreported, and also had complete freedom to direct its analysis and its reporting. There was no editorial direction or censorship from the employer of the authors.

Acknowledgments

Ms Jana Bednarz performed analysis and interpretation of data for this study and is acknowledged with thanks.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data may be made available by contacting the corresponding author.

Additional information

Funding

The authors received no specific funding for this work.