Abstract
Recent studies suggest that personality traits can change through interventions, but little is known about what drives individual differences in intervention gains. Based on personality development and intervention frameworks, we examined whether achieved personality state changes explain differences in trait change and whether these could be explained by participants’ motivation and intervention engagement. The study was based on a 12-week personality intervention with a total of 956 participants, daily and weekly state measures, and trait self- and observer-reports. Participants who showed stronger state deviations from the initial trait level changed more in their traits in the corresponding direction. Both state and trait changes were stronger if participants were more committed to their change goal, completed more implementation intentions, and enjoyed their implementations more (i.e., reinforcement). In contrast, the strength of the desire to change, expected attainability, and beliefs in the changeability of personality had no consistent effects—but individual differences were limited as most participants scored highly on these motivational aspects. The control group exhibited no mean-level changes or effects of any of these covariates. Taken together, these findings support key components of volitional change theories and can help guide future personalized intervention efforts.
Plain language summary
This study explored whether people can intentionally change aspects of their personality and what makes some people more successful at doing so than others. Nine-hundred and fifty-six people took part in a 12-week smartphone guided training where they chose a personality trait they wanted to change, such as becoming more outgoing or more emotionally stable. Throughout the training, they practiced activities aimed at self-reflection and building new habits and completed daily and weekly questionnaires on their thoughts, feelings, and behaviors. We wanted to examine whether changes in one’s behaviors or thought patterns—in other words leaving one’s “comfort zone”—would help people achieve their change goals. We also studied which motivational factors were important for the training to work. We found that people who acted differently than usual (e.g., more outgoing than they typically are) on a regular basis also reported lasting changes in their corresponding personality trait (e.g., Extraversion). Regarding the motivational factors, we found that people who were more committed to changing, enjoyed the activities more, and completed more activities also reported the strongest changes. Simply wanting to change or believing change was possible wasn’t enough on its own—commitment and actual engagement mattered more. These insights could help design more personalized and effective programs to support personal growth.
Keywords
Introduction
One of the most recent observations in personality science is that personality traits can be changed through interventions (Allemand & Flückiger, 2022; Hudson et al., 2019; Roberts, Luo, et al., 2017; Stieger et al., 2021; see Haehner et al., 2024 for an overview). Moreover, recent (volitional) personality change theories (Hennecke et al., 2014; Roberts, 2018; Roberts & Jackson, 2008; Wrzus & Roberts, 2017) and conceptual intervention frameworks (Allemand et al., 2024; Allemand & Flückiger, 2017; Chapman et al., 2014; Magidson et al., 2014; Rebele et al., 2021; Roberts, Hill, & Davis, 2017; Sauer-Zavala et al., 2017) suggest various factors that contribute to or are requirements for volitional personality change, such as the desire to change, beliefs in the changeability of personality, and a prolonged engagement in behavioral change activities. However, most personality intervention studies so far have looked at average treatment effects, and it is unclear which factors of motivation and intervention engagement contribute to the degree of change. The first goal of this preregistered study was to examine whether self- and observer-reported personality trait changes during a 12-week intervention could be explained by preceding state-level changes. The second goal was to study whether the degree of trait and state changes could be explained by eight different motivational and engagement factors suggested in the literature. Specifically, whether people were motivated to change, thought it was possible to do so, adhered to the intervention, enjoyed the intervention activities, and chose particularly difficult tasks.
The link between changes in personality states and traits
Personality traits are typically defined as relatively enduring, automatic patterns of behaviors, thoughts, and feelings, as opposed to their state expressions in behaviors, thoughts, and feelings at specific times and in specific situations (Baumert et al., 2017; Roberts & Jackson, 2008). State expressions of traits are more transient and arise in response to specific situations and contexts. By definition, state expressions deviate above and below one’s central tendency. Theories of personality trait change propose several important factors linking deviations in states to personality trait change. First, if the modal state required by situations and environments is consistent with one’s modal or trait level, then little or no change in the trait should occur (Roberts, Hill, & Davis, 2017; Wrzus & Roberts, 2017). Second, for personality trait change to occur, the state expression called for by situations must be different than one’s modal level. In other words, if one’s job calls for extraverted behavior, and introvert could experience change in the trait of extraversion. Third, for shifts in traits to reflect meaningful change, the shift in states must not only contradict prior modal levels of the trait, but do so over a long enough time for the shift to become ingrained and automatic (Allemand & Flückiger, 2017; Chapman et al., 2014; Hennecke et al., 2014; Magidson et al., 2014; Roberts & Jackson, 2008; Wrzus & Roberts, 2017). For example, the TESSERA (Triggering situations, Expectancy, States/State expressions, and Reactions; Wrzus & Roberts, 2017) framework suggests that repeated short-term, situational processes contribute to long-term personality development. According to this framework, repeated behaviors, thoughts, and feelings triggered by situations, as well as intentions and change goals, can lead to personality development through reflections on these states, habit formation, or conditional learning.
Recent conceptual personality intervention frameworks also endorse this bottom-up approach, in which personality interventions target specific and narrowly defined state expressions of traits rather than acting directly on the broad traits (Allemand & Flückiger, 2017; Chapman et al., 2014; Magidson et al., 2014; Roberts, Hill, & Davis, 2017). This is due to states being more changeable and specific compared to the broad and relatively stable traits, making for an easier intervention target. By using goal-directed activities (e.g., behavioral challenges; emotional regulation exercises) states could be changed in the desired direction until the corresponding traits adapt. These activities are supposed to work best if they are specific (Hudson & Fraley, 2015) and personally important, enjoyable, and in accordance with individual values (Roberts, Hill, & Davis, 2017).
Figure 1 (an adaptation of the illustration of cognitive plasticity in Lövdén et al., 2010) provides an example of the potential relationship between states and traits during an intervention. Without external pressure, states fluctuate around the trait level (i.e., dynamic equilibrium or “comfort-zone” in Figure 1). Through the engagement in intervention activities, the states should start deviating more strongly from the initial trait level in the desired direction (e.g., acting more gregarious than one would typically be) (i.e., prolonged mismatch or state-trait discrepancy in Figure 1). Given that participants maintain the high state level (e.g., keep acting gregarious), the trait should slowly adapt until a new dynamic equilibrium or “comfort zone” is reached. As the trait level adjusts over time, maintaining the new behaviors, thoughts or emotional patterns should become less effortful or more “natural”—whereas considerable effort needs to be invested at the beginning of the intervention. Illustration of potential state-trait dynamics before, during and after an intervention. Note. Adapted from Lövdén et al. (2010).
Although recent theoretical models suggest that prolonged state changes can lead to trait changes, there is still a lack of empirical evidence for this link. For example, as an examination of the TESSERA framework (Wrzus & Roberts, 2017), Quintus and colleagues (2021) examined the association between Big Five states measured at the end of the day for 10 days every two months with trait changes measured every six months. However, they did not find any significant associations between the state level and trait change in most cases. A potential reason for the lack of associations could be that they examined whether people who showed higher state levels would change positively in the corresponding trait—without checking whether the states deviated from their usual level, or in other words whether people acted differently than they usually would. The individual differences in the state levels might just been a reflection of the individual differences in the trait levels (e.g., Whole Trait Theory, Fleeson, 2001; Fleeson & Gallagher, 2009; see also positive initial trait and state correlations in Quintus et al., 2021). Acting in line with one’s trait level should not lead to changes, but only if these states deviate systematically from it (see state-trait mismatch in Figure 1). The state levels might have been high because of an already high trait level, thus not leading to further changes. Other studies with a more specific focus found evidence for correlated change between states and traits, more specifically that increases in sociable behavior were linked to increases in Extraversion (van Zalk et al., 2020) or that changes in momentary stress reactivity were concurrently correlated with changes in Neuroticism (Wrzus et al., 2021). While these studies show that state changes are associated with trait changes, these observational studies cannot clarify whether the traits changed due changes in behavior, or whether the state changes were just a reflection of trait changes due to some unobserved cause. Taken together, the current evidence for associations between state and trait changes is rather weak, and this link has not been examined in the context of personality interventions. To better understand the effects of personality interventions and to further develop evidence-based interventions, the first goal of this study is to examine whether intervention-related trait changes can be explained by the state-trait discrepancy during the intervention.
The role of motivational and engagement factors for state and trait change
One of the most fundamental observations in intervention science is that people often differ in how they respond to and change because of psychological interventions. In psychotherapy, for example, different psychological approaches may produce comparable outcomes and the same treatments can have different effects on the way people change (Delgadillo et al., 2016; Wampold & Flückiger, 2023). Likewise, people differ in the way they engage in their therapy (e.g., Constantino et al., 2018). In theories or frameworks on (volitional) personality trait change, several factors contributing to individual differences in change have been suggested (Allemand et al., 2024; Allemand & Flückiger, 2017; Chapman et al., 2014; Hennecke et al., 2014; Hudson, 2021; Magidson et al., 2014; Roberts & Jackson, 2008; Wrzus & Roberts, 2017). For instance, people should only change or change more if they want to, if they believe it is possible, if they adjust their behavior, and if these changes are reinforced. To better understand potential differential effects and personalize future psychological interventions, it is relevant to examine for whom the interventions are particularly suitable. In this study, we thus focused on the role of several of the proposed motivational and engagement factors as predictors of trait and state change (see Figure 2 for a conceptual model). In essence, we expect participants with a stronger 1) initial motivation (i.e., desire and commitment to change) and 2) higher expected chances of success (i.e., expected attainability and beliefs about the changeability of personality traits) to engage in more intervention activities. Those who 3) engage in more activities overall should show stronger state (and consequently trait) change, in particular if they 4) enjoy these activities more (i.e., reinforcement) and 5) complete particularly difficult activities. Furthermore, a higher enjoyment should encourage participants to engage in more activities afterward. For difficulty, direct effects on the subsequent intervention engagement are less clear: On one hand, managing to complete difficult activities might increase participants’ self-efficacy (Bandura, 1997), further motivating them to engage in more difficult activities. On the other hand, overly difficult activities that exceed one’s capabilities and require considerable effort might backfire and decrease engagement (e.g., Pavlov et al., 2023), emphasizing the need to consider curvilinear difficulty in the analysis. In the following, we discuss these factors in more detail. Conceptual model of individual differences in intervention-related personality change. Note. Based on theories and frameworks of (intervention-related) personality change (e.g., Allemand & Flückiger, 2017; Hennecke et al., 2014; Hudson, 2021; Roberts, Hill, & Davis, 2017; Wrzus & Roberts, 2017). Notable exceptions not covered in this study are intervention properties (e.g., activity specificity; Hudson, 2021) and moderators of state to trait transfer (e.g., reflection on states; Wrzus & Roberts, 2017). All effects are positive unless otherwise specified.
Motivation to change: Change goal strength and commitment
Arguably, people should be more (or only) successful at changing their traits if they wish to do so (Allemand & Flückiger, 2017; Hennecke et al., 2014; Hudson & Fraley, 2015). In line with this, the desire to change has often been suggested as the first prerequisite of volitional trait change (e.g., Hennecke et al., 2014). While there are generally quite some differences in the direction in which people want to change, the majority of people report a desire to change on at least one trait in studies on this topic (Baranski et al., 2021; Hudson & Fraley, 2016; Hudson & Roberts, 2014; Sun & Goodwin, 2020; Thielmann & de Vries, 2021). Regarding the direction, people generally want to change the traits they score particularly high or low on (e.g., to increase Conscientiousness if they have low Conscientiousness levels) or for which they experience dissatisfaction with related life domains (Hudson & Fraley, 2016; Hudson & Roberts, 2014; Olaru et al., 2023; Stieger et al., 2020). For instance, people who are dissatisfied with social aspects of their life generally wanted to become more extraverted, likely to find new friends or a partner to alleviate some of their experienced loneliness.
There is increasing evidence that the change goal strength, or in other words the degree to which people want to change a trait, is associated with longitudinal changes in the corresponding traits over time—although the effects are generally weak (see Haehner et al., 2024 for an overview). For instance, Hudson and colleagues (2020) conducted a mega analysis with a combined sample of 2238 participants and found that people with a stronger desire to change increased slightly in the corresponding traits over the course of the next months (i.e., 0.04 to 0.18 SDs across 4 months; average 0.14 SDs across 3 months in the overview by Haehner et al., 2024). On average, people seem to proactively work towards reaching their desired traits. The effects of change goal strength were also positive but weak in interventions using implementation intentions (Hudson & Fraley, 2015) or weekly behavioral challenges (Hudson, 2023).
While people might have the general desire to change some aspect about themselves, they also need to commit to this goal to enact the necessary changes to reach it (e.g., Heckhausen, 2007). Leaving one’s comfort zone and making the necessary behavioral, affective or cognitive changes to achieve one’s change goals is a difficult task requiring considerable effort. For instance, dispositional introverts may feel more fatigued by having to act extraverted for prolonged amounts of time (Jacques-Hamilton et al., 2018). We thus think that considering the motivation alone without volition is not sufficient to explain potential trait changes. To our knowledge, the commitment to change has not been explicitly assessed so far in personality change studies. However, a recent intervention study showed that merely accepting behavioral challenges without completing them, or in other words wanting to change but not being committed to it, did not lead to trait changes (Hudson et al., 2019). Only those participants who also completed the challenges reported some trait changes in the desired direction.
Expectations of success: Expected attainability and beliefs about the changeability of personality
In addition to wanting to change, theories on volitional trait change suggest that the success thereof is dependent on people’s expectations about the feasibility to change (Allemand & Flückiger, 2017; Dweck, 2008; Hennecke et al., 2014). Here, we distinguish between two components that may affect these expectations, namely whether people think that personality can be changed or is set in stone (Dweck, 2008), also known as growth (vs. fixed) mindset, and whether they think that they would have the opportunities and resources (e.g., support) to enact changes. Motivational theories, such as the expectancy-value (Eccles & Wigfield, 2002) or self-efficacy theory (Bandura, 1997) suggest that higher expectations about one’s capability to achieve specific goals lead to a stronger motivational effort and persistence, particularly in contexts requiring sustained behavioral change. If people think it’s unrealistic to change—either because they think personality is unchangeable or that they do not have the time and resources to do so—they are likely less willing to invest the time and resources needed to change their behaviors for long enough (see precondition 2 of self-regulated personality development in Hennecke et al., 2014). Apart from motivational aspects, the beliefs in the changeability of personality might also affect whether people internalize changes in their states (i.e., behaviors, thoughts, and emotions) as part of their personality self-concept (see reflective processes in the TESSERA framework of personality change; Wrzus & Roberts, 2017). People who think that their personality cannot change might be more likely to attribute their achieved behavioral changes to the intervention activities instead of changes in their underlying dispositions and may continue to seek out familiar situations and habits once they discontinue the intervention activities.
A large body of research has shown that a stronger self-efficacy or better outcome expectations are linked to positive health behavioral change (e.g., Ashford et al., 2010; Rhodes & Dickau, 2012; Sheeran et al., 2017; Strecher et al., 1986) or academic achievement (e.g., Honicke & Broadbent, 2016; Richardson et al., 2012; Robbins et al., 2015), but we know of no study that has examined the effects on volitional trait change. Similarly, studies on growth mindset (i.e., that core attributes like intelligence and personality can change) have predominately focused on academic behaviors or academic achievement and generally found weak positive effects (for an overview see, Burnette et al., 2022; Yeager & Dweck, 2020). In the context of personality change, Hudson and colleagues (2021) examined associations between the beliefs in the changeability of personality and subsequent trait change across 4 months and found no effects.
Engagement in personality change activities
After a realistic goal has been set and committed to, the next step of volitional trait change should be to make the necessary behavioral changes in the desired direction (e.g., Hennecke et al., 2014; Roberts, Hill, & Davis, 2017). In the context of personality interventions, participants can be supported in modifying their behavioral, affective, and thought patterns with challenges (e.g., “ask a friend to coffee”), implementation intentions (e.g., “if it is Monday at 9:00, I will go jogging”), instructional feedback (i.e., coaching) or self-reflection exercises (e.g., reflecting on consequences of current and desired trait level in specific situations) (Hudson & Fraley, 2015; Massey-Abernathy & Robinson, 2021; Stieger et al., 2021). The more participants engage in these activities, the stronger should be the state and consequently trait changes they can achieve. Hudson and colleagues (2019) found support for this hypothesis by showing that the number of completed behavioral challenges during a 15-week timespan predicted trait change in the desired direction. These findings were replicated in a later study using a similar design (Hudson, 2023).
Reinforcement of change
Solely changing one’s behavior on a few occasions is likely not enough to elicit any meaningful trait change, but it needs to be maintained for a long enough time until the changes become habitual or automatized (e.g., Allemand & Flückiger, 2017; Hennecke et al., 2014; Wrzus & Roberts, 2017). While it might be possible to achieve this through constant repetition of behaviors, it is unlikely that people will keep engaging in novel behaviors if they do not deem them worthwhile, or in other words, if the behaviors are not reinforced (Wrzus & Roberts, 2017). Reinforcement (and punishment) of behaviors, either through one’s own emotional reactions or through feedback from the social environment, has been commonly suggested as one of the main ways through which traits are shaped by the environment (Caspi & Roberts, 2001; Roberts & Wood, 2006; Wrzus & Roberts, 2017). The same processes are expected to apply to interventions (Allemand & Flückiger, 2017; Chapman et al., 2014; Wrzus & Roberts, 2017). If people feel that the changes made are useful, are valued by others, or increase their momentary well-being, they should be more likely to repeat these changes and consequently achieve longer lasting habit and trait changes. A lack of rewards, or even punishment (e.g., negative affect; criticism from others), on the other hand, should reduce attempts to change early on, or require a stronger commitment and effort to be overcome.
Experimental studies provide evidence that conditioning can lead to behavioral changes (e.g., Delgado et al., 2009; Houben et al., 2010). Assessing momentary behaviors (i.e., social context and physical activity) and affect in daily life with an experience sampling study, Wichers and colleagues (2015) were able to show that a more positive or negative affect lead to an increased or decreased likelihood of engaging in similar behaviors at a later time point, respectively (but see Heininga et al., 2017 for no replication of the effects). Little is known about whether these effects would also accumulate into later trait change. Quintus and colleagues (2021) examined whether daily affective experiences across 50 days—coupled with the state expressions during the same day—were associated with longitudinal changes on the corresponding Big Five traits. The hypothesis was that high state levels combined with a strong positive affect would lead to increases in the trait, whereas the opposite would be the case for a combination of low state levels and positive affect. However, they did not find any evidence for this. Taken together, most of the evidence for the effects of reinforcement on shaping behavior comes from experimental studies so far, with weak support from observational studies.
Activity difficulty
While not explicitly mentioned in the intervention or volitional change frameworks, we expect the difficulty of the activities to also explain individual differences in change. If participants engage in more difficult activities, they should deviate more strongly from the “comfort zone” or trait-typical behaviors (Allemand & Flückiger, 2017; Gallagher et al., 2010). This in turn might have a stronger impact on personality change. Furthermore, successfully completing challenging activities might increase one’s self-efficacy and expected chances of achieving one’s goal (Bandura, 1997; Eccles & Wigfield, 2002). A competing perspective is that activities of moderate difficulty provide the best balance between demands and individual capabilities (e.g., flow theory; Csikszentmihalyi, 1990). In the context of academic achievement, difficult learning tasks are associated with less engagement and lower learning rates than easier ones (e.g., Pavlov et al., 2023). Hudson and colleagues (2019) examined whether the difficulty of chosen behavioral challenges during a personality intervention was associated with the degree of trait change, but found no significant effects.
The present study
The main goal of this pre-registered study was to examine factors contributing to individual differences in personality state and trait changes during a 12-week digital personality intervention (Stieger et al., 2018, 2021). More specifically, we first wanted to examine how daily and weekly states change over the course of the intervention, and whether these are related to self- and observer-reported trait change. We then investigated whether individual differences in state and trait change were explained by motivational and engagement factors suggested in theories on (volitional) personality change (e.g., Allemand & Flückiger, 2017; Hennecke et al., 2014; Wrzus & Roberts, 2017), namely the 1) change goal strength, 2) change goal commitment, 3) expected attainability of their change goal, 4) beliefs about the changeability of personality, 5) number of implementation intentions completed, 6) number of opportunities identified to practice the implementation intentions, 7) experienced valence of implementation intentions, and 8) difficulty of implementation intentions. Based on the theoretical propositions, we expected all of these to be positively associated with state and trait change. For most of these suggested predictors of change this study represents the first empirical investigation in the context of a personality intervention, with previous studies only providing some initial support for the effects of the change goal strength (e.g., Hudson & Fraley, 2015) and completion of intervention activities (e.g., Hudson et al., 2019).
Methods
The hypotheses and data analysis were pre-registered: https://osf.io/8mht3 (for deviations see Open Science Statement). Data, analysis code, and supplementary materials are available at https://osf.io/9str3. This research was conducted according to the Declaration of Helsinki and the full study protocol was approved by the Ethics Committee of the University of Zurich. We report how we determined our sample size, all data exclusions (if any), all manipulations, and all measures in the study.
Participants and procedure
We used data from the PEACH intervention study (Stieger et al., 2018, 2021). Participants were recruited in Switzerland via university mailing lists, social media advertisements, flyers, or word-of-mouth. The participants had to be 1) 18 years or older, 2) able to read German, 3) not in a psychotherapeutic or psychiatric treatment, 4) owner of a smartphone (Android or iOS) with mobile internet connection, and 5) interested and motivated to participate in the intervention and to change a personality trait. Adults with mental health disorders and other psychosocial problems were screened out and provided with contact information to the psychological counseling service at the University of Zurich.
At the beginning of the study, participants were asked to choose one out of nine change goals (i.e., a combination of a Big Five trait domains and increase vs. decrease; except for decrease in Emotional Stability) from a list describing each change goal (Stieger et al., 2020) based on the items of the Big Five Inventory 2 (Soto & John, 2017). After completing the pre-test, one third of participants was randomly assigned to a 4-week passive waitlist control group, whereas the other two thirds started with the self-administered 12-week intervention. As the only incentive participants received for this study was access to the intervention app, we kept the waitlist phase relatively short to ensure that control group participants would still be motivated to participate. The control group did not differ significantly from the intervention group in any of the variables considered here at the pre-test (i.e., absolute differences between the groups ranged from d = 0.01 on the initial trait level to d = 0.07 on the expected attainability of success).
The intervention included several types of intervention activities: 1) self-generated behavioral implementation intentions (e.g., “If I have to work concentrated, then I switch into flight mode”), 2) pre-defined weekly behavioral challenges (from which participants choose one; e.g., “Tidy up a part of your flat every day”), 3) psychoeducation (e.g., short knowledge clips or messages about personality traits), 4) self-reflection activities (e.g., observing situational triggers that help or hinder desired behaviors), and 5) resource activation (e.g., gratitude diary; identifying personal strengths) (for a detailed overview see Stieger et al., 2018). Control group participants could start the intervention after the waitlist phase. In addition to a post-test at the end of the intervention and waitlist phase, participants completed daily (only intervention phase) and weekly questionnaires. They were also asked to contact at least three close friends, family members and/or romantic partners to provide observer-reports for the personality traits at the pre- and post-test. Because we were interested in individual differences in change, we only included those participants who provided at least two repeated measures. Overall, this resulted in 956 participants, or more specifically 552 for self-reported traits, 281 for observer-reports, 693 for the weekly measures, 819 1 for the daily measures, and 229 participants from the control group (i.e., who completed both the control group pre- and post-test).
The participants (511 women; 53.4%) were 18 to 69 years old (M = 25.45; SD = 7.39). The majority of the sample consisted of students (55.0%). A total of 652 (43.4%) participants reported to be currently working part- or full-time. With regard to highest completed education, 45.4% of participants reported to have completed a university entrance qualification, 21.0% a Bachelor’s degree, 16.3% a Master’s degree, 7.2% vocational training, and 4.8% a university of applied sciences degree. We did not assess ethnicity in this study, but most participants reported German as their mother tongue (90.3%) and to be of Swiss (83.7%) or German (9.3%) nationality. Compared to the general population of (the German-speaking parts of) Switzerland, the sample was thus generally younger and more highly educated. With regard to the chosen change goal, the majority of participants wanted to either increase their Emotional Stability (28.1%), Conscientiousness (26.7%) or Extraversion (22.3%), followed by much smaller groups wanting to increase in Open-mindedness (7.0%), decrease in Agreeableness (6.4%) or increase therein (4.3%).
Measures
Personality measures
Self-reported (SR) traits
At the beginning and end of the intervention (12 weeks later), as well as beginning and end of the control group waitlist phase (4 weeks later), participants were asked to fill out the German adaptation of the Big Five Inventory-2 (BFI-2; Danner et al., 2019; Soto & John, 2016). The BFI-2 is a 60-item measure of the Big Five personality trait domains and 15 corresponding facets (i.e., three facets per trait domain). Each facet is measured by four items, two of which are negatively keyed. Participants responded to the items on a 5-point Likert scale ranging from 1 = strongly disagree to 5 = strongly agree. McDonald’s ω ranged from .82 for Agreeableness to .89 for Extraversion and Conscientiousness.
Observer-reported (OR) traits
At the beginning and end of the intervention (12 weeks later; no additional OR measure in the waitlist phase), informants recruited by the participants filled out the German adaptation of the observer-report Big Five Inventory-2 short version (BFI-2S; Soto & John, 2017). The BFI-2S is a 30-item measure of the Big Five personality traits and 15 facets. Each facet is measured by two items, one of which is negatively keyed. Participants responded to the items on a 5-point Likert scale ranging from 1 = strongly disagree to 5 = strongly agree. We only used observer-reports if they were provided by the same observer on both measurement occasions. McDonald’s ω ranged from .76 for Open-mindedness to .87 for Conscientiousness and Negative Emotionality.
Weekly personality states
At the end of each week during the intervention or waitlist phase, participants were asked to fill out the German Big Five Inventory-2 short version (BFI-2S; Soto & John, 2017). The BFI-2S is a 30-item measure of the Big Five personality traits. To focus on participants’ personality states during the last week, the instruction (i.e., “Please rate the extent to which the following statements apply to yourself in the last week”) and items (e.g., “I was rather quiet”) were adjusted. Participants responded to the items on a 5-point Likert scale ranging from 1 = strongly disagree to 5 = strongly agree. This questionnaire was administered 3 times during the waitlist phase and 11 times during the intervention (i.e., end of first week to last week before the post-test). On average, the included participants completed 1.83 (SD = 1.09) waitlist and 7.04 (SD = 3.48) intervention phase assessments. McDonald’s ω ranged from .75 for Agreeableness to .85 for Negative Emotionality. The intra-class correlation was .92 at the aggregate level (which we used for the main analyses) and .51 for the single measures.
Daily personality states
At the end of each day during the intervention, participants were asked to report their daily states (“How do you rate yourself today?”) on five bipolar items reflecting the Big Five (i.e., Extraversion: “shy vs. gregarious”; Emotional Stability: “unsure vs. confident”; Conscientiousness: “rash vs. deliberate”; Open-mindedness: ”narrow-minded vs. open-minded”; Agreeableness: “dismissive vs. empathetic”). Participants indicated their state-level with a slider ranging from 1 to 100. This questionnaire was administered 70 times starting at the end of the first week (day 7) to the end of the last week before the post-test (day 76). On average, included participants completed 27.65 (SD = 21.10) assessments. The intra-class correlation was .98 at the aggregate level and .42 for the single measures.
The length of the assessments differed because we strived for a balance between assessment quality and participant burden. We chose a shorter version for the weekly assessments because of the higher frequency thereof, as well as for the observers as we expected them to be less motivated to fill out longer scales (as compared to the participants themselves). For the daily states we used a very short measure to ensure that participants would stay motivated to fill out the daily measures throughout the intervention.
We did not examine each trait (or change goal group) separately but merged all nine into one. We did so to increase power, and because we did not have any trait-specific hypotheses. For each participant, we thus only used the responses on the measure for the trait they wanted to change (e.g., Extraversion scores for participants wanting to increase Extraversion, Agreeableness scores for participants wanting to increase Agreeableness) and combined all of them into one variable. For participants wanting to decrease a trait, we first reversed coded the scores (e.g., recoding Extraversion scores into Introversion scores for participants wanting to decrease Extraversion). Higher scores on this combined variable should thus represent higher levels on the targeted trait. We did this for all four personality measures (i.e., self-reported traits, observer-reported traits, weekly states, and daily states).
Personality change predictors measured at pre-test
For all predictors for which this was possible, we measured them at the pre-test. The rationale for this was that we wanted to exclude the possibility that they change because of the (lack of) intervention changes, thus affecting the associations with the trait or state change. For example, participants who experience stronger trait changes might start believing more strongly in the changeability of personality, inflating the associations if it was measured at the end of the intervention as compared to the beginning.
Change goal strength
The desire to change was assessed at the beginning of the study with a single item (i.e., “Please indicate how much you want to change according to your change goal.”) on a scale ranging from 0 = not at all to 7 = very much.
Change goal commitment and attainability
Change goal commitment and attainability was measured at the beginning of the study with an adaptation of the 6-item scale developed by Brunstein (1999) to measure personal goal commitment and attainment. Three items measured commitment (e.g., “Under no circumstances am I prepared to give up on my change goal.”) and three attainment (e.g., “I have enough opportunities in my everyday life to put my change goal into action.”). Participants responded on a scale ranging from 1 = strongly disagree to 5 = strongly agree. We used the mean scores for the analyses. Because the scale was adapted for this context, we first examined if the expected structure would hold. A parallel analysis suggested two components. An exploratory factor analysis with maximum likelihood extraction and oblimin rotation found that the three commitment items loaded by λ = .50 to .83 on the first factor, whereas the three attainment items loaded by λ =.38 to .58 on the second factor. Cross-loadings were all close to 0 (i.e., |λ| < .05). The two factors correlated with each other by r = .50. McDonald’s ω was .72 for commitment and .47 for attainability.
Beliefs about the changeability of personality
The belief about the changeability of personality was measured at the beginning of the study with the 8-item implicit theory of personality scale (Dweck, 2013; e.g., “You can always substantially change the kind of person you are”) on a scale from 1 = not true at all to 5 = very true. Four items were reverse coded so that higher scores indicate a stronger belief about the changeability of personality. We used a mean score across the eight items for the subsequent analyses. McDonald’s ω for the scale was .89.
Personality change predictors measured daily during the intervention
The following predictors all relate to the implementation intentions and were thus measured during the intervention, more specifically every evening during the intervention—except for Sundays when participants formulated new intentions for the upcoming week. We measured these on a daily basis to ensure that participants still remembered their experiences as well as possible.
Implementation intention completion and identified opportunities
Participants were asked every evening (except Sundays) whether they fulfilled their implementation intentions (0 = no; 1 = yes). They were also asked to report how many opportunities they identified to practice their implementation intention on, regardless of whether they used the opportunity or not. For the number of completed implementation intentions we used the sum score across all days. We also did so for the number of identified opportunities.
Implementation intention valence (reinforcement)
As an operationalization of reinforcement, we assessed the positive affect that participants experienced when completing the implementation intention. When participants reported to have completed their implementation intention, they were asked how pleasant it was for them (from 1 = very unpleasant to 7 = very pleasant). We used the mean score across days as an indicator of the average experienced difficulty and valence. Because the effects of implementation intention and valence may accumulate (e.g., several somewhat positive experiences having a stronger affect than a single very positive one), we also multiplied the z-standardized valence with the overall number of fulfilled implementations. This score should thus represent the accumulation positive experiences or reinforcement throughout the intervention.
Implementation intention difficulty
Similarly to valence, the difficulty of implementation intentions was only measured when participants also reported to have completed them. It was measured on a 7-point Likert scale ranging from 1 = very easy to 7 = very difficult. Similarly to valence, we used the average reported value across days as covariate, and also considered the product between z-standardized difficulty and number of completed implementation intentions. The rationale was that several somewhat difficult activities should have a stronger impact than a single very difficult one. To test whether a moderate difficulty would be ideal, we also computed the squared centered difficulty and reversed the values so that a moderate difficulty would have the highest, and a very low or very high difficulty the lowest scores.
Statistical analysis
State deviation (State change)
To obtain the state deviation, we extracted the residuals (εj) from a linear regression in which the average state level was the dependent variable and the initial trait level the independent variable
We did so because it is not a high state level per se that should be important for trait change, but a strong deviation from the previous state or trait level (see Figure 1). Because we did not have a pre-test state measure as a reference, we used the initial trait level. For the sake of simplicity, we refer to this as state change in the following.
Trait change
For the main analyses, we used multilevel models in which the measurement occasions were nested within individuals (i.e., pre- and post-test, weeks or days). To estimate trait change across the intervention, we added time as a predictor to the multilevel models using trait scores as dependent variables (0 = pre-test; 1 = post-test). For the first research question (is state and trait change associated?) we also added the state deviation (from the initial trait level) as a predictor. A positive interaction effect (β3) between time and the state deviation would suggest that people who changed their states more strongly during the intervention also changed more on the traits
For the remaining RQs, we again predicted the trait scores by time and the covariates. Again, a positive interaction (β3) would suggest that people with, for example, stronger change goal strength would change more
For the states, we also used a multilevel model and predicted the states by the covariates and added the pre-test trait scores as a control variable—again to estimate the deviation from the trait level. A positive main effect of the covariates (β2) would thus suggest that people with, for example, a stronger change goal strength showed larger state changes
We ran the multilevel models separately for each covariate, because some predictors are likely mediators of other effects (e.g., completed implementation intentions as a mediator of the effect of change goal strength, commitment and attainability). Including all predictors and thus controlling for the effects of the other predictors, would result in partial effects (e.g., the effect of commitment apart from implementation intention completion), which provide an underestimation of the total effects. All covariates were z-standardized based on the entire sample for the analyses. All analyses were run in R version 4.5.0 (R Core Team, 2025) with the R packages ggplot2 (Wickham, 2016), haven (Wickham et al., 2025), lme4 (Bates et al., 2015), and psych (Revelle, 2025).
Results
Descriptives and correlations
Descriptives and correlations between personality measures and predictors.
Note. II = Implementation intentions. Traits and states are based on the first measure (i.e., pre-test for traits or end of first week for states). */**p < .05/.01.
The correlations between the predictors, as well as initial personality measures and predictors, were generally very low. The only exception was a very high overlap between the number of implementation intentions completed and opportunities identified to use them on (r = .83), and between commitment and the change goal strength (r = .46).
State and trait change
Trait and state means across the measurement occasions are presented in Figure 3. Unstandardized means (transformed to a 1 to 5 scale for the daily states) are presented in the upper panel. The standardized difference to the first measurement is presented in the bottom panel. While the initial SR trait levels and weekly states showed similar mean levels at the beginning and an increase over the course of the intervention, the daily states were rated consistently high throughout the intervention (M = 3.43 with a range of 3.31 to 3.53; transformed from 0 to 100 to a 1 to 5 scale for the sake of comparison). Except for the daily states, all other measures indicated some mean-level increase during the intervention, but with largely varying effect sizes. The largest gains (i.e., difference between last and first measure, tested with paired t-tests) were found for the weekly state measures (d = 0.67; p < .001), followed by the SR traits (d = 0.31; p < .001), and OR traits (d = 0.07; p = .012). The control group reported no significant mean-level changes during the waitlist phase (SR traits: d = −0.06; p = .18; weekly states: d = 0.12; p = .23). Unstandardized and standardized trait and state means across time. Note. The upper panel represents unstandardized means (transformed to a 1 to 5 scale for the daily states), the lower panel the standardized difference to the first measure (i.e., pre-test for traits; end of first week for weekly and daily states). Error bars represent bootstrapped 95% confidence intervals. SR / OR = Self- / Observer-report; CG = control group.
Associations between states and traits.
Note. WS/DS = Weekly/Daily states; SR/OR = Self-/Observer-reported traits; CG = Control group.
Predictors of state and trait change
Next, we added the motivational and engagement factors as predictors of the state or trait change. The associations are presented in Figure 4. Aside from the observer-reports, the findings were very consistent across measures. Most notably, people reported stronger trait or state changes when they were more committed to change (β= .07 to .10), completed more implementation intentions (β = .11 to .18; excluding OR), and enjoyed them more (β = .10 to .15 or β = .16 to .19 when taking frequency into account; excluding OR). For the other predictors, the findings were less consistent. The change goal strength was only significantly associated with SR trait change (β = .10). Expected attainability had no significant effects. Stronger beliefs in the changeability of personality resulted in stronger gains in the observer-reports (β = .09) and daily states (β = .08). And finally, easier implementation intentions were associated with state changes in the desired direction (WS/DS β = −.11). We also examined if a moderate difficulty would have the strongest effects, but none of the associations with the squared centered difficulty (including interactions with the number of implementation intentions) were significant. In the control group, none of the covariates explained individual differences in change. Associations with trait and state change. Note. SR / WS = self-reported traits / weekly states; II = Implementation intention; valence/difficulty x number = interaction effect between standardized valence or (linear) difficulty and number of completed implementation intentions. Presented are standardized regression weights with 95% confidence intervals. Solid black lines indicate significance (i.e., the confidence interval does not include 0).
Exploratory analysis: Predictors of momentary state change
Because our prior analyses relied on aggregates of the implementation intention measures across the intervention (e.g., total number of implementation intentions completed, average difficulty, and valence) we wanted to check if these associations also hold at the momentary level, or in other words if the states would be higher on those days or weeks in which participants completed their implementation intentions and found them less difficult or more pleasant. To do so, we used the multilevel models to predict the daily or weekly states by the implementation intentions completion, opportunities, difficulty and valence, for example
Concurrent effects of implementation intentions on states.
Note. II = Implementation intention. β = standardized effects (based on z-standardized variables). The models were run separately for each predictor, except for the two difficulties (linear and squared centered), which were run together. Significant effects (i.e., with the confidence interval not including 0) are marked in bold.
Discussion
Recent studies have suggested that personality traits can be changed through intentional change efforts (e.g., Allemand & Flückiger, 2022; Hudson et al., 2019; Roberts, Hill, & Davis, 2017; Stieger et al., 2021). However, psychological interventions are not equally successful for everyone, and intervention effects can be characterized by large individual differences. The goal of this study was thus to better understand which factors contribute to personality trait change or individual differences in personality interventions.
In summary, we found that those people who were able to change their states most strongly compared to their initial trait level also changed most strongly in the self- and observer-reported trait levels. Furthermore, participants showed stronger trait or state changes in line with their goals if they 1) were more committed to their change goal, 2) completed more implementation intentions, and 3) enjoyed them more (i.e., were reinforced). We found no consistent support for the other factors. More specifically, the change goal strength only predicted a stronger self-reported trait change, but neither observer-reports nor state change. Beliefs in the changeability of personality were associated with stronger observer-reported trait change and daily state change, but not the other two measures. Identified opportunities for practice seemed much less relevant than the actual implementation intention completion. Easier implementation intentions were associated with state levels closer to the change goal, but not trait change. And finally, the expected attainment showed no associations with change.
Trait and state change
Before going into details about individual differences in change, we wanted to discuss some discrepancies between the mean-level changes across the four sources of data, namely the very weak increases for the observer-reports and no mean-level changes for the daily states. This does not seem to be due to a general disagreement across the measures, as the levels or initial scores were associated by β/r = .43 to .66 (see Tables 1 and 2), ranging towards the upper end of commonly found associations for self-and observer-reports (Connelly & Ones, 2010; Connolly et al., 2007) or state and trait measures (Augustine & Larsen, 2012; Fleeson & Gallagher, 2009; Kritzler et al., 2020). More likely, these represent differences in what people and observers interpret as trait change, or how participants respond to state vs. trait measures.
First, the self-observer discrepancy may be due to an over-exaggeration of trait-changes by the participants, or because observers did not have enough time or trait-related interactions to update their pre-existing view on the target’s personality. While we found some associations between the states and observer-reported changes, these were considerably weaker than the associations with self-reported trait change. In this study, participants primarily recruited their romantic partner, family members or close friends as observers. For future intervention studies it might be more beneficial to select observers who 1) know the participants for a shorter time period, 2) have frequent interactions with the participants shortly before and during the intervention, and 3) have interactions that are relevant to the chosen change goal. For example, when trying to increase their Conscientiousness levels, participants could primarily ask colleagues to provide observer-reports instead of parents or siblings, which might have fewer interactions during the intervention with a focus more on other aspects.
Second, the daily states showed no increase over the course of the intervention. Participants might have also adjusted what they would consider a high state-level over the course of the intervention, for example, feeling very extraverted when talking to a stranger at the beginning of the intervention, but less so towards the end of the intervention. Another likely reason is that the daily state measures started at the end of the first week of the intervention, at a point at which participants were already engaging with the intervention. On the first assessment day (i.e., day 7 of the intervention) the daily state mean was already at M = 60.05 (on a slider from 0 to 100 with a starting value of 50) and remained around this level throughout the intervention. Given that participants generally wanted to change those traits they had low levels on (M = 2.84 on the pre-test trait measure; see also Hudson & Fraley, 2016; Hudson & Roberts, 2014; Stieger et al., 2020), the state levels should have been lower under non-intervention conditions. Because 82.5% of the daily states were reported on days in which the implementation intentions were completed, these were consequently much higher than usual (i.e., a difference of 7.30 in the daily states between completion and non-completions days). To overcome this, we used the pre-test trait level as an anchor to compare the states against. In future studies, establishing a state baseline prior to the intervention would be desirable to estimate state changes more precisely.
Despite these limitations, the state level deviations from the trait level were associated with stronger trait changes, both in the self- and observer-reports. While some studies examining the link between states and trait change focused on the average state levels (e.g., Quintus et al., 2021), we expected that a deviation from the trait level or usual behavior should be the relevant driver of trait change (e.g., Allemand & Flückiger, 2017; Chapman et al., 2014; Hennecke et al., 2014; Wrzus & Roberts, 2017). In line with this, participants who deviated most strongly from their initial trait level showed more pronounced trait changes, both in the self- (weekly/daily β = .46/.27) and observer-reports (weekly states β = .10). These findings provide evidence in favor of the proposition that leaving one’s comfort zone of typical behaviors, emotional reactions or through patterns can lead to pronounced trait changes.
The state deviations were also associated several of the motivational or engagement factors (i.e., change goal commitment; beliefs in the changeability of personality; implementation intention completion, valence, and difficulty). These patterns mirrored the findings for the self-reported trait change, as well as the observer-reported change associations to some degree, indicating that similar motivational and engagement processes can trigger a state and consequently trait change. In the following, we discuss these findings in more detail.
The role of motivational and engagement factors for state and trait changes
Motivation to change: Change goal strength and commitment
Arguably the first prerequisite to intentional personality trait change is that people want to change some aspect of their personality or behavior (Allemand & Flückiger, 2017; Hennecke et al., 2014). In our study, participants’ change goal strength at the beginning of the study was weakly associated with their self-reported trait change, but not with the observer-report or state measures. In contrast, the commitment to change predicted self- and observer-reported trait change as well as the weekly and daily states. Taken together, these findings might suggest that the volition to change is more important than the general desire to change, although people who wanted to change more strongly also seemed to be more committed to their goal (r = .42 in this study).
While the effect sizes were small, they might provide an underestimation of the true effects due to several design decision for this study or personality intervention in general. First, we assessed change goals at the broad trait domain level. However, participants might have only wanted to change specific aspects thereof and not the entire trait domain (e.g., gregariousness instead of Extraversion; Olaru et al., 2024). When change goals and traits are assessed at the facet level, associations should be stronger than for trait domain ratings (Hudson & Fraley, 2015). Furthermore, the previously found effects of the change goal strength in intervention studies (Hudson, 2023; Hudson & Fraley, 2015), including this one, do not seem to exceed those in observational studies (e.g., Hudson et al., 2020). This is surprising given that the intervention studies provided participants with structured activities to support them in achieving their change goals. In personality intervention studies, participants self-select into change goal groups, which substantially decreases the variance in the desire to change compared to observational studies based on large heterogeneous samples. For instance, the reported change goal strength in our study was generally very high, with 75% of participants reporting a value of 5 or higher on a scale of 0 = not at all to 7 = very much. Similarly, 50% reported a commitment value of 4 or higher on a scale from 1 to 5. When also recruiting participants who do not want to change and participate for other reasons (e.g., course credits or financial incentives), we expect the effects on trait change to be much more pronounced. However, this would also not provide an adequate representation of people who would engage in volitional personality change efforts in the first place, and result in an underestimation of the average treatment effect.
Expectations of success: Expected attainability and beliefs about the changeability of personality
For the desire to change to be translated into actual behavioral or trait changes, we expected that participants should believe in the possibility to change (Allemand & Flückiger, 2017; Hennecke et al., 2014). In line with this, we found the belief in the changeability of personality to be positively associated with the daily state and observer-reported trait change, although the effects were generally weak and not significant for self-reported traits and weekly states. Hudson and colleagues (2021)—using the same measure of beliefs in the changeability of personality (or personality growth mindset; Dweck, 2013)—did not find any link to self-reported trait change over the course of several months. Studies on growth mindset—typically in an educational context—have shown that it is associated with higher motivation (Burnette et al., 2013), but that the effects of mindset interventions on more objective outcomes (e.g., academic achievement) are very weak (Burnette et al., 2022). Furthermore, a recent psychoeducation-focused personality intervention with regular newsletters about the relevance of personality traits and the possibility to change them (Baranski et al., 2025) did not show any effects on trait change—or even the motivation to change one’s traits. Taken together, the effects of a growth mindset (intervention) alone seem limited, although there might be some effects if integrated with other intervention approaches (e.g., behavioral activation).
Regarding the expectations about available opportunities and resources to practice (i.e., attainment), we did not find any effects. While the attainability scale had very low internal consistency due to the different components covered (ω = .47), it still correlated to a similar extent with the other variables as for instance the commitment scale (e.g., both were moderately correlated with the experienced valence; see Table 1). As such, the lack of associations was likely not just due to a lower reliability of the scale. The findings might thus suggest that the expectations for opportunities to practice do not matter, at least in the form in which they were measured here. An issue limiting conclusions about this is that our sample represented a selection of motivated people who likely already expected to have time to change, with 66% reporting an attainment value of 4 or higher on a scale from 1 to 5. In contrast, those people with lower expectations about their prospective success might have been less likely to join without substantial financial incentives.
Engagement in personality change activities
We used several intervention activities in this study, including psychoeducation, self-reflection exercises, resource activation tasks, behavioral challenges, and implementation intentions. However, we only assessed the adherence to the implementation intentions. Nonetheless, we found that the number of implementation intentions completed was associated with a stronger state and self-reported trait change (β = .07 to .13). Furthermore, participants reported states in line with their change goal on days in which they completed their implementation intentions (β = .16 or B = 7.30 on the 0 to 100 slider). Implementation intentions or behavioral challenges represent a commonly used bottom-up intervention activity (Hudson et al., 2019; Hudson & Fraley, 2015; Massey-Abernathy & Robinson, 2021), with interventions using these activities also finding significant self-reported mean-level changes in the targeted traits. We contribute to this literature by furthermore showing a dose-response association between the number of completed implementation intentions and state or trait change (see also Hudson, 2023; Hudson et al., 2019 for association between number of completed behavioral challenges and trait change). Interestingly, the number of opportunities people identified to practice their implementation intentions on did not have consistent effects on the state and trait change. When holding the number of practice opportunities equal, it thus seems preferable for them to be spread out across several occasions instead of focused around a few moments (i.e., days or weeks) during the intervention.
Reinforcement of change
Behavioral changes are thought to lead to a stronger trait change if they are reinforced (Allemand & Flückiger, 2017; Caspi & Roberts, 2001; Chapman et al., 2014; Roberts & Wood, 2006; Wrzus & Roberts, 2017), or in other words if participants experienced these changes as more positive. In line with this hypothesis, we found the valence of the implementation intentions to have some of the strongest effects on the self-reported trait and state change in this study (when taking number of completed activities into account: β = .10 to .16 with self-reported trait/weekly state/daily state change). The daily valence experienced when completing the implementation intentions was also positively associated with within-person state changes that day (see Table 2), although it is unclear whether more positive experiences lead to stronger state expressions or whether achieving state changes in the desired direction triggered a more positive affect (for effects of achieving trait change on well-being see e.g., Hudson & Fraley, 2016; Olaru et al., 2023). While reinforcement (or punishment) has been shown to be able to lead to changes in human behavior (e.g., Delgado et al., 2009; Houben et al., 2010; Wichers et al., 2015; but see Heininga et al., 2017) the presumption that reinforced states lead to trait change has not been supported so far (e.g., Quintus et al., 2021). In this study, we focused on reinforcement through the emotional experience, but future studies could also consider feedback from others. Social interactions, in particular the reaction of others to one’s behaviors, serve as relevant sources of reinforcement or punishment (see e.g., Back et al., 2011). An interesting question for future research could be how the effects of social reinforcement compare to those of internal emotional reactions, and whether the former could explain differences in observer-reported trait changes.
Activity difficulty
We additionally considered the experienced difficulty of enacting the behavioral changes, with the expectation that completing more difficult intentions would lead to a stronger state or trait change. In contrast to this, the difficulty was not liked to trait change, and easier implementation intentions were associated with states in the desired direction. However, it is unclear whether participants were able to enact stronger state changes when they chose easier implementation intentions, or whether they rated them as less difficult when they were better able to express their desired state level. As such, we do not know to what extent the reported difficulty can serve as an objective metric of the actual difficulty of the tasks. Because the implementation intentions were created by participants themselves, we did not have alternative metrics of difficulty to investigate this in more detail. For instance, Hudson and colleagues (2019) used a behavioral activation intervention in which participants were presented with a pre-specified list of behavioral challenges, with the difficulty of each being rated by the authors. However, they also did not find any associations between trait change and the average difficulty of the completed challenges—only with the number of completed challenges. Even if the difficulty of the activities does not matter for trait change, using an adaptive difficulty approach when suggesting activities for the next days or week (see Hudson et al., 2019) might be useful to ensure that participants stay motivated throughout the intervention (Fong et al., 2015).
Limitations and future directions
There are several strengths of this study, such as the use of four different measures of personality change on different time scales, the wide range of motivational and engagement factors considered, and the large sample size. However, we want to point out some limitations that can guide future research. First, the findings of this study may not be generalizable to the general population, but only to people who want to change an aspect of their personality and have the time and resources to commit to this change effort. Naturally, not everyone is motivated to change an aspect of their personality or to do so with the help of a digital support application (Allemand & Stieger, 2024). This selective sample also restricted the variance on many of the motivational factors measured here (e.g., change goal strength, commitment, attainability), and consequently our ability to find associations with these measures. While recruiting participants who do not want to change or assigning them randomly to change goal groups may improve the estimation of these associations due to increasing captured individual differences, it would also result in an underestimation of the average treatment effect for people who would engage in volitional personality change efforts in the first place. We also think that asking people to engage in activities that should change their behavior irrespective of their own change goals is unethical and does not align with the idea of providing a tool to help people make changes towards their own ideal version of themselves.
Second, we did not assess why people wanted to change these aspects of their personality. While previous research has shown that the desire to change a trait is stronger if people are less satisfied with related aspects of their life (e.g., desire to become more conscientious if satisfaction with work is low; Hudson & Roberts, 2014; Olaru et al., 2023), assessing the precise reasons or the extent to which this motivation is intrinsic (e.g., desire to feel better) or extrinsic (e.g., societal or cultural norms; pressure from partner or work; comparisons with others) might be a valuable predictor to include in future studies (Deci & Ryan, 2000). Doing so would be particularly worthwhile in other cultural contexts or cross-cultural studies, in which the effectiveness of personality interventions may differ because of different reasons to change and contextual factors supporting or hindering change efforts.
Third, the daily and weekly states were only measured once the intervention started. We used the initial trait level as a proxy in this study, but future studies should include a pre-intervention state assessment as a better baseline (see Figure 1). In addition, the daily states were only measured with a single item and once every evening. Longer state measures would allow for a broader construct coverage and the ability to examine whether participants generally focus on changing only specific behaviors or several states related to the targeted trait domain. Several momentary state measures each day would further improve the quality of the assessment and allow researchers to pinpoint the moments in which the intervention activities were completed, and whether their effects are momentary or also have lagged effects across longer time spans. By further including situational measures, future studies could also examine changes in situation-state contingencies (e.g., Kuper et al., 2022) during the intervention, such as weaker emotional reactions to stressors during an Emotional Stability intervention. However, we deemed such a detailed assessment to be too demanding for participants during the 12-week intervention phase. Similarly, observer-reports of participants’ weekly or daily states during the intervention would be ideal to see if others detect some changes in the behavior of the targets (Jayawickreme et al., 2022), and most importantly under what circumstances they use this information to update their perception of the target’s personality. This would also provide an indication of the number of interactions between the participants and their observers, as well as whether these interactions were related to the change goal. A less demanding way to assess behavioral changes in the naturalistic behavior during the intervention could be to include mobile sensing approaches (e.g., Beierle et al., 2018; Mehl, 2017; Rüegger et al., 2020; Stachl et al., 2020).
Fourth, we only measured adherence to the implementation intentions but not the other intervention activities (e.g., self-reflection exercises, psychoeducation, behavioral challenges). Doing so would allow future studies to also compare the various activities against each other and disentangle the unique effects of each.
And finally, we focused on general patterns of change throughout the intervention (but see Table 3). An alternative way of examining trait or state change—if sufficient repeated measures are available—would be to look at individual trajectories in change, in particular at non-linear trajectories and sudden shifts (e.g., Lambert et al., 2003). In psychotherapy research, sudden gains during the treatment have been shown to be associated with overall better treatment outcomes (Lutz et al., 2013), whereas the opposite was the case for sudden losses (i.e., a decrease in therapy progress). In personality interventions, this could potentially be examined by looking at discontinuous shifts in the state or trait levels of participants or based on repeated measures of the subjectively experienced change goal progress. Interesting questions for future research would be whether such sudden changes in the behavioral, emotional, or thought patterns of participants are associated with differences in the overall trait change, and which intervention processes trigger these shifts. Tracking the intervention progress in more detail and adjusting activities or incorporating face-to-face sessions when deemed necessary (e.g., when progress is slow or decreasing but participants still report to want to change) could furthermore help improve the attainment rate and overall outcomes of personality interventions (see e.g., Lambert et al., 2003).
Conclusion
The current study makes several important contributions to the research on intervention-related personality change. We found that participants who showed more pronounced state changes from their initial trait level, or in other words left their comfort zone, also changed more in their trait levels—both in their own and observers’ assessments thereof. Furthermore, this study suggests that intervention-related state and trait changes are stronger when participants are more committed to their change goal, complete more implementation intentions, and have more positive experiences (i.e., reinforcement) when doing so. For the other factors, we found less or no support. The strength of the desire to change was only associated with stronger self-reported trait change, although our ability to find effects was limited by a generally very strong desire to change in this sample. The beliefs in the changeability of personality only had very weak effects on the observer-reported trait and daily state change. We found no support for the hypotheses that the expected attainability of the change goals and the difficulty of the activities would be associated with intervention gains. In summary, this study provides evidence for most of the suggested drivers of volitional personality change (Allemand & Flückiger, 2017; Hennecke et al., 2014; Wrzus & Roberts, 2017) and provides some guidelines for future personality-intervention research.
Footnotes
Acknowledgment
This article is based on data from an interdisciplinary trial funded by a grant from the Swiss National Science Foundation (No. 162724; PI: Allemand). We would like to thank our colleagues Tobias Kowatsch and Dominik Rüegger for their collaboration on the project and for programming the app-based intervention. We thank Sara Aeschlimann, Chantal Gerl, Lara Keller, Marcel Lauber, Elias Laimer, Marcia Nissen, Fabienne Thierstein, Moritz Truninger, and Nadia Wohlwend for their help in preparing study materials and collecting data.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This article is based on data from an interdisciplinary trial funded by a grant from the Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (No. 162724; PI: Mathias Allemand).
Open science statement
The hypotheses and data analysis were pre-registered (https://osf.io/8mht3). Data, analysis code, and Supplemental Materials are available at the OSF repository and can be accessed at https://osf.io/9str3. In the analyses, we deviated from the pre-registration in the following ways: We also included the control group as part of the revision. We modeled state change as the discrepancy to the initial trait level as opposed to linear change during the intervention (see OSF for results based on the preregistered analysis). We originally assumed that the states would increase slowly over the course of the intervention, similar to previous studies that used weekly trait measures (e.g., Hudson et al., 2019). However, participants already reported very high daily state levels from the first measure on, and the linear state change for both daily and weekly states were not associated with any other of the variables included here (e.g., with the number of completed implementation intentions). We used multi-level models instead of latent growth curve models because these are more common and computationally efficient for a large number of repeated measures than structural equation models (see Chou et al., 1998). The results are equivalent, independent of which method we used to estimate the trait change and individual differences therein (see OSF for findings based on latent growth curve models). Regarding the measures, we used the initial instead of weekly change goal strength. We originally planned to use the weekly measure to have several assessments thereof. However, as participants got closer to their change goal, their desire to change decreased (d = −1.05 across the intervention; p < .001). As such, the average weekly change goal strength was negatively confounded with the achieved changes, which was not the case for the pre-test measure. We moved the findings for general preference for behavioral practice and self-reflection to the OSF because of the length of the study and less of a conceptual support from (volitional) change theories. And finally, we did not include the number of days participants completed the daily measure, which was intended as a measure of general intervention engagement as we did not track the actual app log-ins. However, participants primarily filled out daily measures on days when they completed implementation intentions (on 82.5% of days with responses), making this measure redundant with implementation intention completion (r = .89).
Ethical considerations
This research was conducted according to the Declaration of Helsinki and the full study protocol was approved by the Ethics Committee of the University of Zurich.
Consent to participate
Participants gave written consent for participating prior to the start of this study.
Supplemental Material
The preregistration of the hypotheses and analyses can be found at https://osf.io/8mht3. Data, analyses code, and Supplemental Materials are available at the OSF repository and can be accessed at 
.