Materials and method

A systematic literature review allows us to contrast, analyze and synthesize the findings contained in the literature with methodological rigor. More specifically, this research is a systematic review with a psychometric focus, seeing as it promotes the analysis of the psychometric properties of instruments, which makes it possible to examine reliability and validity indices (^{Munn et al., 2018}). In this research we analyzed the measurement instruments of a single construct: personal agency. We used the protocol for systematic reviews based on PRISMA, with steps that include the identification, screening, eligibility and inclusion of documents (^{Page et al., 2021}).

During the stage of search and identification of research corpus, the databases listed were Scopus, Web of Science, PsycINFO, Medline, SciELO and IndexPsi, the first four being international bases and the last two national. The searches were carried out on October 21, 2020, updated on February 14, 2021 by two reviewers who used the strategy: (“sense of agency” OR “human agency” OR “judgment of agency” OR “agency judgment” OR “personal agency” OR “types of agency” OR “forms of agency” OR “perceived agency” OR “feelings of agency”) AND (measur* OR assessment OR validity OR evaluation OR inventory OR scale OR “measures of agency” OR instrument)), searched in titles, abstracts and keywords with the exception of SciELO, in which we searched only by title and abstract due to the limitation of the search platform. The two national databases included the same terms in Portuguese and Spanish.

As search filters, we limited the results to documents in Portuguese, English and Spanish, with the exception of PsycINFO, where it was not possible to include the language filter. In addition to the databases, we used the references of the articles already found to seek other documents to attach to the research corpus. At the end of this stage, we downloaded the databases and the screening stage began. We removed duplicate data and two independent reviewers began eligibility analysis by reading the abstracts, considering the following inclusion criteria: a) approach the concept of agency from the cognitive and behavioral perspective; b) quantitative or mixed studies; c) mention of psychometric instruments that measure personal agency; d) be in Portuguese, English or Spanish and available for online access.

The last stage of data collection refers to the extraction of data from documents, seeking information about the instruments used, dimensionality, number of items, population and sample, and psychometric properties. During data analysis, we used the COSMIN checklist (Consensus-based Standards for the selection of health status Measurement Instruments), a tool that assists in evaluation of methodological quality in research on the measurement properties of instruments (^{Mokkink et al., 2010}). For this study, we used only the items related to the identification of measurement properties, which are: internal consistency, reliability, measurement error, content validity, internal structure validity, hypothesis testing, cross-cultural validity, criterion validity and responsiveness, verification of convergent, divergent and construct validity.

Results

Several scales set out to measure agency as a factor of other constructs, but there are few specific instruments for measuring this phenomenon. Of the 3.451 documents identified, after organizing and filtering the database by reading the abstracts with the inclusion criteria in mind, only 377 abstracts were kept for the full reading stage to be analyzed as to their adequacy to the research objective. There were 17 of them which could not be verified, since they were not available in their full versions; we identified 141 documents with proposals for measuring personal agency, however, only 42 of them had instruments for the specific purpose of measuring this construct in general.

In the bibliographic references of these documents, we identified 15 other studies dealing with the development of psychometric measures of agency. Only 13 of them were included because two were not available, totaling 55 selected documents. The flowchart of data collection and delimitation of the research corpus is illustrated in Figure 1.

Figure 1: Flowchart of data collection and delimitation of the research corpus 

Among the 55 documents which use tools that meet the inclusion criteria for the research, we identified a total of 23 instruments intended to measure personal agency. Some of these documents refer to the construction or adaptation and validation of instruments, while others apply these tools in their research. The tools identified are described in Table 1 8, 9,10,11, which contains information about their dimensions, items, research sample and psychometric properties.

Table 1: Instruments for measuring personal agency and its psychometric properties 

Based on Table 1, we can observe that 21 of the 23 instruments analyzed provide analysis of internal consistency and reliability. In 20 of them Cronbach's alpha was used, which is one of the most used procedures to measure the internal consistency or reliability of an instrument (^{Zanon & Hauck Filho, 2015}). The most widely used validity criterion for this group of instruments is the internal structure validity (or factorial, internal, and construct validity), measured on 15 instruments in different ways. This type of validity indicates whether the responses obtained by the instrument are internally consistent with respect to the different parts that make up the tool (International Test Commission, 2014; ^{Pasquali, 2007}). Hypothesis testing, which refers to verifying the acceptance or rejection of initial hypotheses formulated by researchers, was identified in eight instruments. This measurement property is fundamental for obtaining construct validity (^{Souza et al., 2017}).

Convergent validity was identified in six instruments and makes it possible to verify that the indicators of different phenomena have a significant relationship, in which both have high variance in common (^{Hair Jr. et al., 2019}). It can be obtained by correlating instruments that measure a similar phenomenon, in which a high correlation between the factors is expected. In the case of divergent validity, only four instruments provided analyzes that present evidence of this type of validity, which is observed through the use of the measure of a construct opposed or in the opposite direction than the one being investigated (^{Polit, 2015}). In addition to divergent validity, content validity was mentioned in four instruments. Its importance is to enable a definition of the nature of the phenomenon to be measured and provide important information about the construct validity (^{Vianna, 1983}).

The measurement error is indicated by four instruments, two of which indicate the identified standard error -the EAPI and the PAS. This data can be identified by observing the variation in the scores obtained between the applications of an instrument in the same participant, which is caused by the variation in error scores. Thus, it refers to the standard deviation of error scores (^{Primi, 2012}). The measurement error is indicated by the standard measurement error. Such data indicates that the true measure referring to an attribute is located between the average value of the measurements performed and a standard error both up and down (^{Pasquali, 2003}).

We found criterion validity in four instruments. This type of validity can be predictive -it makes it possible to identify to what extent the instrument is able to predict future behaviors based on the evaluation at a given time or criterion-; researchers compare the results obtained through the scale with the results of another scale regarding the same construct considered “gold standard” (^{Pacico & Hutz, 2015}; ^{Polit, 2015}). Cross-cultural validity was verified in only two instruments and was partially performed in four others, since not all of them present all the steps indicated by ^{Beaton et al. (2000}). The information we found relating to the properties of measures in the 23 instruments identified are summarized in Table 2.

Table 2: Summary of the measurement properties of the instruments that measure agency 

^Note: * Process carried out partially or incompletely.

Through the synthesis proposed by Table 2, it should be noted that no instrument demonstrated the responsiveness of the instrument. It is also possible to verify that not all the tools presented different types of validity, and five instruments present conceptual and statistical designs that were not very deep. Such elements were not found in the documents that present their development or were not available.

Discussion

The identified instruments are analyzed from three prisms: a) theoretical perspectives and historicity of the instruments; b) psychometric properties and; c) population and sample of the studies which developed the instruments and languages that are available.

Theoretical perspectives and historicity of instruments

Different theoretical perspectives regarding studies on agency can be observed through the measurement instruments of the construct from the theoretical framework that supports them and are reflected in the dimensionality of the scales. The first instruments identified are aligned with the theoretical perspective that relates sex, gender and the influence on agency and communality, which is based on the idea of representation of social roles of men and women and the associated stereotypes. From this understanding, the sense of agency is associated with the masculine, with characteristics of determination, individualism, action, competence, ambition, dominance, and objectivity, for example. The sense of communality, on the other hand, is directed to the feminine, contemplating care, collectivism, morality, interdependence and parental investment, for example (^{Abele et al., 2008}). We identified nine instruments in alignment with this perspective: PAQ, EPAQ, GEPAQ, IAT, GRI-14, AC, ACJ, AC and ASMS.

The other theoretical perspective observed in the instruments is associated with the triadic reciprocal causation model and metacognitive processes. This model assumes that environmental factors, personal factors, and motivation for action and behavior influence one other in agency capacity. Metacognition plays a fundamental role in this process, in which the main resources are intentionality, premeditation, self-reactivity and self-reflection (^{Bandura, 2001}, 2018). Fourteen instruments are associated with this perspective: BIF, APAB, CAMI, Multi-CAM, MAPS, MAPS-20, PIAS, EAPI, PAS, ESAGE, AHA, AQ, PAB and ATPA-22.

Even with 14 instruments developed, which consider distinct dimensionalities related to the triadic reciprocal causation model and metacognitive processes, no instrument was identified that addresses all the agency factors identified by Gai et al. (in press). The authors propose an expanded dimensionality of personal agency from the bandurian perspective, which include: intentionality, premeditation, self-reactivity, self-reflection, self-efficacy and internal control. This model contemplates the metacognitive processes indicated by ^{Bandura (2001}), as well as self-efficacy (Bandura, 1989) and internal control proposed by ^{Rotter (1966}). This indication of a possible limitation regarding the theoretical basis in the construction of these instruments, which is reflected in the dimensionality of the scales.

Regarding the temporal distribution of the development of these scales, different movements were observed regarding the development of instruments from different theoretical perspectives (Figure 2).

Figure 2: Distribution of the development of the scales of each theoretical perspective by year 

Based on Figure 2, we can observe that the instruments that refer to agency and commonality and gender stereotypes are older and recommend tooling for measuring the construct. However, they have less consistency regarding the development of new scales, since as after the 1981 scale new instruments were only identified in 2006, and we found no tools have been made since 2015. Conversely, even with the first instrument related to this perspective dating back to 1989, the theoretical line that associates agency and the triadic reciprocal causation model and the metacognitive processes presents greater consistency in the development of new instruments.

The reason for the decrease in the use of instruments associated with the notion of agency and gender stereotypes may be associated with the demystification of this stereotype, especially since they are related to impairments in emotional development and maturation, leading to psychological distress (^{Jesus et al., 2020}). The influence of interpersonal relationships on the perception of agency is amplified by ^{Landes and Settersten Jr. (2019}), who criticize the lack of recognition of this aspect in the literature, even though it is encompassed in the conceptualization of Bandura as part of external influencers. The authors understand that human agency is dependent on relational aspects and propose that the conceptualization and measurement of agency require the consideration that people's lives are interconnected, and that the perception of total “independence” is flawed, contradicting the notion of antagonism between agency and communality.

Conclusion

From the execution of this systematic literature review in which we sought to analyze the psychometric properties of instruments for measuring personal agency, we identified 23 instruments that propose to measure personal agency in a specific way. Of the 55 documents analyzed, some are specifically aimed at the construction or adaptation and validation of these instruments, while others presented empirical research using these tools.

Two groups of instruments were observed based on the theoretical support that underpins them, identified in the aforementioned literature and the factors that make up the scales. The first instruments are based on the perception of agency and commonality, derived from gender stereotypes. The second strand of instruments is aligned with the triadic reciprocal causation model and metacognitive processes, which contemplate the greatest number of tools. We identified no instruments that concomitantly contemplate the expanded dimensionality of agency proposed by Gai et al. (in press): intentionality, premeditation, self-reactivity, self-reflection, self-efficacy and internal control, something that emphasizes the need to build new instruments.

There are weaknesses regarding the robustness of the psychometric properties of the instruments, since most are limited to the provision of validity evidence of internal structure. Convergent and divergent validities are little explored, which may be due to the difficulty in identifying instruments that enable this verification, or due to the lack of conceptual clarity about agency in order to establish these criteria. We identified no instruments for measuring agency with statistical validity for Portuguese, nor in the Brazilian context. The only instrument adapted for this language presented only semantic validation -the Agency Beliefs (^{Smith et al., 2000}) adapted by ^{Fontes et al. (2010}).

From this data, we consider that the development of an instrument adapted to the Brazilian context is the main recommendation of studies from this systematic review. The lack of an instrument that meets the validity criteria hinders the advances in the literature on agency in the country. The construction of an instrument that contemplates the expanded dimensionality of agency proposed by Gai et al. (in press) would make it possible to fill this theoretical gap. Such an instrument can be an option for researchers who have been studying the subject and end up using instruments to measure related constructs, as in the case of the study by ^{Dressler et al. (2019}).

Among the limitations of this study, even with six databases being screened, it is still possible that other instruments were left out, due to the difficulty of covering all available studies and also because documents of gray literature were not considered. Additionally, we studied documents from up until the year 2020, and some of the documents that developed instruments could not be accessed, which limited the description of the development of the tool and its psychometric properties. As suggestions for future studies, it is worth recommending research development in languages other than English, since, out of the 23 instruments studied, only six were not in this language. This may be one of the possible justifications for the small number of studies about agency capacity in Latin America, including Brazil.