A review of the types of mobile activities in mobile inquiry-based learning

Inquiry-based Learning is increasingly suggested as an efficient approach for fostering learners’ curiosity and motivation. It helps learners to develop their ability to work on complex and unpredictable environments making them more critical thinkers and agentic learners. Although, mobile technology is a suitable support for this learning process, there is a lack of practical strategies for educational practitioners to enact the right balance between enabling agency and supporting the students through the mobile technology. Thus, we conducted a literature review that analyzed 62 studies on mobile inquiry-based learning. The analysis focused on the level of agency supported by mobile technology. This review study provided two main results. The first result is a two-layer classification –with five types and twelve subtypes– of the most common mobile activities used in inquiry-based learning. The types and subtypes are: 1) Direct instruction formed by 1a) location guidance, 1b) procedural guidance and 1c) metacognitive guidance, 2) Access to content formed by 2a) fixed and 2b) dynamic content, 3) Data collection that consists of 3a) cooperative and 3b) collaborative data collection, 4) Peer-to-peer communication formed by 4a) asynchronous and 4b) synchronous social communications and 5) Contextual support that includes 5a) augmented experience, 5b) immersive experience and 5c) adaptive feedback. The second result consists of an analytical framework –based on six dimensions– to assess the level of agency supported by the different types of mobile activities. The learners’ agency dimensions are: 1) Goals, 2) Content, 3) Actions, 4) Strategies, 5) Reflection and 6) Monitoring. Finally, the review presents insights on how this analytical framework can be used by educational practitioners to identify mobile activities that effectively balance learners’ agency with mobile technology.


Introduction
The term inquiry-based learning (IBL) has been defined in multiple ways in the literature. Originally, it referred to a continuous exploration of a topic of students' interest, in which learners engage in social interactions to generate shared understanding (Piaget, 1959;Pierce, 1955;Vygotsky, 1978). De Jong & van Joolingen (1998) defined it as an educational strategy based on discovering knowledge that fosters active participation and learners' responsibility. Pedaste & Sarapuu (2006) referred to inquiry-based learning as an approach in which learners solve problems by using their inquiry skills. Recently, inquirybased learning has been increasingly suggested as an efficient approach for fostering students' curiosity and motivation by linking science teaching in schools with informal learning and phenomena in everyday life (Specht, Bedek, Duval, & Held, 2012). Such strategy can help students to develop their ability to work on unpredictable and complex environments, especially in the current, ever-changing, technology driven, society. In this context, reports like the one from UNESCO (West & Vosloo, 2013) has shown that mobile technology provides opportunities for learners to engage in a broader range of informal learning activities, especially to make learners responsible for executing learning actions. Mobile technology provides opportunities for more personalized and autonomous seamless experiences across learning contexts (Thüs et al., 2012), in which learners have freedom and power to make proactive decisions. Sharples, Taylor, Vavoula (2007) suggested that learners' agency lies in a socio-cultural synergy between all the individuals willing to advance knowing, instead of relying on the individual learner. The associated shift in the locus of control over learning (Chen, Tan, Looi, Zhang, & Seow, 2008;Sharples et al., 2005) may be facilitated by seamless learning designs that distribute agency, to make learners more self-directed.
However, increasing learners' agency and offering less support can lead to less desirable learning outcomes (Sawyer, 2005) or to learners struggling to select, organize and integrate relevant information (Kirschner, Sweller, & Clark, 2006;Mayer, 2004). That is why the scaffolding concept, used by Bruner & Sherwood (1976) as the form and quality of the effective intervention by a 'learned' person in the learning of another person, becomes essential. The division of activities between teachers and students, the required level of teachers' involvement, and the importance of students' autonomy and agency in their own learning remain important concerns to achieve efficient inquiry-based learning processes (Brown & Campione, 1994;Drexler, 2010;Kirschner, Sweller, & Clark, 2006;Klahr & Nigam, 2004;Lumpe & Oliver, 1991;Mayer, 2004;Sha et al., 2012;Song, Wong, & Looi, 2012). The issue of learners' agency in mobile inquiry-based learning has been superficially addressed in Nouri, Cerratto-Pargman, Rossitto, & Ramberg (2014). The study, that compared two inquiry-based teaching approaches with and without mobile technology, put the focus on the different actions that were supported with mobile technology and how those actions had an effect on the degree of learners' agency and control. This review deepens on this analysis of the actions supported with mobile technology in inquiry-based learning and how learners' agency is affected. In the remainder of this manuscript, these actions supported with mobile technology will be called mobile activities. Thus, the objective is to provide more guidance for practitioners to enable an efficient balance of learners' agency and scaffolding using mobile technology for inquiry-based learning.
This review focuses on how mobile activities in inquiry-based learning enable the distribution of agency and the balance of student support. To this end, this review is structured as follows: Section 2 underpins the theoretical background upon which this review builds: self-directed learning and agency in mobile learning and seamless inquiry support. Section 3 presents the classification framework derived, used to analyze the studies. Section 4 summarizes the outline of the methodology used. Section 5 presents the results based on the classification framework. Section 6 elaborates on the results achieved and discusses its possible implications. Last, section 7 provides the conclusion of this review with the limitations of this study and the outline for further research on the field of mobile inquiry-based learning.

Theoretical Background
A considerable amount of review studies has been published lately on inquiry-based learning, mainly focusing on the effect of the inquiry approaches on learning outcomes and guidance. A recent literature study (Alfieri, Brooks, Aldrich, & Tenenbaum, 2011) investigated the effects of guidance in discovery learning approaches comparing unassisted discovery learning versus explicit instruction and enhanced discovery learning versus explicit and unassisted. The comparison showed that unassisted IBL did not benefit learners, while scaffolding and worked-out examples did. These results were aligned with the meta-analysis (Furtak, Seidel, Iverson, & Briggs, 2012) about experimental and quasi-experimental inquiry-based learning studies, in which it was suggested that teacher-led inquiry lessons had a larger M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 3 effect on students' learning than those that were student-led. Another analysis of the literature focused on the structure and phases of the inquiry (Pedaste et al., 2015). The results presented a synthesized framework that proposed a general structure for inquiry-based learning in order to ensure effective implementation of inquiry-based learning processes. More recently, a meta-analysis (Lazonder & Harmsen, 2016) has also confirmed that guidance is pivotal to successful inquiry-based learning. In this study authors compared the effectiveness of different types of guidance for a different range of age. The results demonstrated that learners who received some kind of guidance improved learning outcomes, but the specificity of the guidance received and the age of the learners did not affect the results.
Other more recent reviews have been focusing on inquiry-based learning supported with mobile technology. Zydney & Warner (2016) identified typical designs with mobile apps (technology-based scaffolding, location-aware functionality and digital knowledge sharing and construction) but also suggested some recommendations for better alignment of the learning theories and the measured outcomes. This review determined that basic scientific knowledge or conceptual understanding were the main outcomes measured in the studies, and more effort from researchers was needed to assess high-level cognitive outcomes like cognitive load or problem solving. Another analysis on the effects of mobile technology on inquiry-based learning (Sung, Chang, & Liu, 2016) focused on the shortcomings linked to the duration of interventions, the methods of measurement of higher-level skills, and also on the weak orchestrations of mobile activities in generic learning activities. This supports the idea that student's activity alone is not enough to explain learning outcomes; scaffolding is often distributed in the learning environment, across the curriculum materials or educational software, the teachers or facilitators, and the learners themselves (Puntambekar & Kolodner, 2005). Despite the results of earlier reviews on (mobile) inquiry-based learning, none have yet analyzed the relation and impact of using mobile technology on learners' agency.

Designing seamless inquiry support
Sharples in (Wong, Milrad, & Specht, 2015) defined Seamless Learning as a continuous flow of meaningmaking despite the changes in the physical and the social context. In the literature, different learnercentered approaches can be identified: for instance, contextual learning theories originating from distributed cognition and contextual immersion, the socio-cultural perspectives that stresses the importance of a systemic approach, constructivist approaches that highlight the importance of learner activity, and furthermore, for designing optimal learning experiences the idea of self-directed learning and scaffolding are discussed (Koole, 2009;Sharples, Taylor, & Vavoula, 2010).
In the field of inquiry-based learning, the integration of mobile technology enables new forms of guidance (Land et al., 2011), interactivity or immersion in technology-supported inquiries. Koole (2009) defines aspects of the device, aspects of the learner and social aspects as part of the complex interaction of mobile learning support. Based on her framework balanced and effective learning experiences can be designed by these aspects. In a similar approach, Luckin (2010) defines a framework in which learning takes place in an Ecology of Resources (EoR). She uses the concept of filters to select relevant resources for supporting learning in the current context. Considering the special case of mobile language learning, Kukulska-Hulme (2009) points out that the focus of attention is actually on the pedagogical interactions M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 4 and less on the nature of the technology being mobile or fixed. Taking learning outdoor benefits learning, as it facilitates an active interaction process between the learners and their environment (Fahraeus, 2004;Wilde, Harris, Rogers, & Randell, 2003).
The context-bound nature of learning is also emphasized in socio-cultural approaches. As Säljö (2010) argues, our knowledge is built through our application of external tools (intellectual, physical, or both), which we integrate into the flow of our activities. Learning is mediated by conceptual and physical artifacts and by other people, in an interacting system of elements. Sharples, Taylor, & Vavoula (2010) postulated that knowledge is embodied in the elements of a distributed system and in the interactions between those elements. The capabilities of mobile technology enable learners to hold conversations across multiples contexts and to generate synergies to co-create knowledge. De Jong, Specht, & Koper (2008) provided a reference model for mobile social software, engagement and immersion. In their framework they analyzed the participating stakeholders, embeddedness into the context, underlying pedagogical paradigms and the data flow of different mobile social applications. Park (2011) proposed a framework to classify the pedagogical affordances of mobile designs based on the attributes of mobile devices; the mobility hierarchy defines four levels to cluster mobile devices and their pedagogical affordances as productivity, flexible physical access, capturing and integrating data, and communication and collaboration support.
Taken together, all these learner-centric approaches supported with mobile technology offer suitable learning activities, at the right time and place (Peng, 2009). These different types of mobile activitiesthat will be used later for the classification framework-provide ubiquitous capabilities for learners to receive specific instruction, guidance and content when they need it (Kukulska-Hulme, 2009;Luckin, 2010). Learners can socially interact and co-create knowledge among distributed participants in a community (Koole, 2009;Säljö, 2010;Sharples et al., 2010). They can also back up their investigations digitalizing evidences for later analysis and reflection (Tabuenca, Kalz, Drachsler, & Specht, 2015;Verpoorten, 2012). Mobile activities can also bridge learners and their environment for supporting more augmented experiences (Fahraeus, 2004;De Jong et al., 2008;Wilde et al., 2003). These properties were also mentioned in (Zydney & Warner, 2016) as part of the typical designs featured with mobile apps in inquiry-based learning. However, do these properties really lead to effective seamless learning? This question, that has been raised by Vogel, Kennedy, & Kwok (2009) and shared by Wong, Milrad, & Specht (2015) in their book, suggesting that just using mobile technology for (inquiry) learning does not necessarily lead to meaningful seamless learning processes. Eventually, this also requires learners to be agents of their own learning in a way that they control their behavior and cognition (Wong, Milrad and Specht, 2015). So, the key question is to what extend those learner-centric approaches supported with mobile technology enable learners to take the responsibility on their learning process (learners' agency).

Learners' agency
During the last decades autonomy and agency have been central to education, and they have been used indistinctly to refer to the ability or capacity to take control of one's own learning (Benson, 2007;Holec, 1979;Holec, 1981) or to refer to the capacity to act with initiative (or self-regulation) and effect in one's own learning (Fay, 1996;Lantolf & Pavlenko, 2001). Some authors (Hunter & Cooke, 2007) claimed that the initiative aspect is the differentiator between autonomy and agency. They argued that learners' agency comprises the initiative of the learner, whereas learners can act autonomously even when they have received specific instructions before. All in all, learners' agency assumes learners to be responsible for the purpose of their learning, the content, the rhythm and the strategies they used as well as for monitoring the progress and assessing its outcomes (Little, 2007). Controlling the process of one's own learning is strongly linked to the concept of self-regulated learning. In its definition (Zimmerman, 1990), learners are characterized by having the control over their metacognition, motivation and behavior, which help them to achieve desired learning outcomes. This definition was extended in the socio cognitive theory of (Bandura, 2001;Zimmerman & Schunk, 2001), in which self-regulated learners should be also able to define their own motivational and metacognitive strategies, manage their own learning situations and choose the right amount of instruction needed. Moreover, as part of the social cognitive theory, Bandura (2001) defined agency as the capability of humans to set a goal and act on these choices. Additionally, he provided four main features that characterized human's agency: a) intentionality, that refers to the power to originate actions. b) forethought, that suggests that an agent should anticipate his/her goals and take the appropriate action to achieve them, c) self-reactiveness, that refers to the power to not only plan and think ahead, but also to motivate and self-regulate as well and d) self-reflectiveness, as the capability to subjectively judge their own performance and assimilate also external feedback.
Several critiques in the literature associated to the lack of support and guidance in discovery learning processes (Kirschner et al., 2006;Mayer, 2004) claimed that enabling too much freedom can generate less desirable learning outcomes and make learners struggling to select, organize and integrate relevant information. Such critiques leverage the need to achieve more understanding of the role of mobile technology supporting learners' agency. Although there has been some preliminary attempts to discuss this issue in the context of inquiry-based learning (Nouri et al., 2014), this manuscript aims at providing practical strategies to scrutinize how learners' agency and scaffolding can be better identified when using mobile technology. Taken together, the agency definitions described in this section will be used in the next section to define the dimensions that will characterize learners' agency.

Analytical framework
To ensure the identification of the opportunities for exercising learning agency in inquiry-based learning supported with mobile technology, this review looks at the link between learners' agency and the state-ofthe-art of mobile device and mobile applications found in literature. In order to conduct this research, this review proposes a review framework which is based on the definitions of learners' agency.
Starting from Benson's (2007) and Holec's (1979Holec's ( , 1981 ideas of learners' agency, we developed an analytical framework to analyze learners' agency while using mobile technology in inquiry-based learning. The framework includes six dimensions and it is based on the following four definitions of agency proposed by Bandura (2001): goals, actions (extracted from intentionality feature), strategies (extracted from the forethought feature) and reflection (extracted from the self-reflectiveness and the selfreactiveness features). This framework was complemented with Little's (2007) control, and with Zimmerman's (1990) definition that incorporates monitor control (extracted from the self-regulation theory). The six dimensions of the analytical framework are: • Goals: Does the learner have any opportunity to set up his/her own learning goals? • Action: Does the learner have any opportunity to decide what to do? • Strategy: Does the learner have any responsibility on how to do it? • Reflection: Does the learner have any opportunity to reflect upon the strategies used to execute the process? • Content: Does the learner have any opportunity to decide which information he/she wants to use? • Monitoring: Does the learner have any opportunity to monitor his/her progress? 4. Method

Selection
Following the structure proposed by the PRISMA statement, as a guideline for undertaking systematic reviews (Moher, Liberati, Tetzlaff, & Altman, 2009), the underlying exploration was conducted in April 2016 using the following databases: Elsevier and Web of Science were used because they contain journals that publish research combining educational and technical aspects. ACM and Springer were selected because they contain one of the largest digital libraries for conference proceedings. And finally, Taylor & Francis was used for covering subjects in education, engineering and social science.
The search for articles was done in three iterations. The first iteration was performed using the keywords included in the identification phase (see Figure 1). In this case, we examined the abstracts of the articles looking for studies describing inquiry based learning processes supported with mobile technology. This identification phase left 110 articles that were considered for further analysis. In the screening phase, 5 duplicates were removed. In the eligibility phase, we filtered out 43 articles without practical implementations of mobile inquiry-based learning, i.e. theoretical studies. Since not a lot of experimental studies were found, case studies and semi-experimental were also considered for the analysis. At the end, 62 studies were retained for an in-depth analysis.
We focused on inquiry-based learning, inquiry learning and science inquiry, because we considered these terms to be sufficient for our analysis. In addition, the review covered studies in the timeframe between the appearance of the first smartphone in 2006 until to 2016.

Analysis
Previous review studies on inquiry-based learning (Furtak et al., 2012;Lazonder & Harmsen, 2016;Maaß & Artigue, 2013) and mobile inquiry-based learning (Sung et al., 2016;Zydney & Warner, 2016) have already addressed topics like learning outcomes, strategies for the practical uptake of IBL, effects of using mobile technology for IBL or the effects of guidance. However, the issue of learners' agency has been addressed superficially in few occasions like e.g. in Nouri et al. (2014). Thus, to deepen in the analysis of how the different types of mobile activities affect learners' agency, a three-step process analysis has been carried out.
Following a joint iterative coding approach (Bogdan & Biklen, 1998;Huberman & Miles, 2002), the first step of the analysis was to cluster all types of mobile activities found in the studies. For that, an initial classification of types of mobile activities has been created based on the ubiquitous pedagogical capabilities of mobile technology described in section 2.1. A ubiquitous pedagogical capability is a unique affordance that mobile technology provides to support learning. The initial types were: 1) direct instruction, 2) access to content, 3) data collection, 4) enable communication and 5) contextual support. At the end of this first step, these types were refined and became the final 12 types of mobile activities supported in inquiry-based learning (see Appendix A). The second step consisted of the development of the analytical framework presented in section 3. Out of the existing definitions of learners' agency, 6 dimensions were developed in order to understand to what extent a type of mobile activity supports or M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 8 limits agency. In the last step of the analysis, the 12 types of mobile activities have been analyzed according to the 6 dimensions of learners' agency that focus on the learners' control over goals, content, actions, strategies, reflection and monitoring. To decide whether a type of mobile activity complies with a given agency dimension, we looked in the studies for evidences that explain the signs of agency. A type of mobile activity complies with an agency dimension when at least one example -explaining how this agency dimension is supported-is found in the studies. Conversely, if a type of mobile activity does not comply with an agency dimension can be due to two reasons: either for a given type of mobile activities there were no evidences in the studies that support the agency dimension, or there were signs that the agency dimension was limited in at least one of the studies.

Results
In this review, 62 studies were selected for further analysis. Out of the initial grouping phase, the distribution of types of mobile activities over the 5 major categories (see Appendix A) was the following: fostering teachers' direct instruction (44/62 studies -70%), access to content (43/62 studies -69%), supporting data collection (57/62 studies -91%), facilitating social interaction (20/62 studies -32%) and enabling interactions between the context and the learner (16/62 studies -25%). In this section, examples of the different types of mobile activities are given together with the analysis of the learners' agency based on the 6 agency dimensions. Thus, two tables will be included in each sub-section; the first table contains examples of mobile activities, the number of times that they were found in the different studies and which technological affordances of the mobile devices were used. The second table contains the scores of each type of mobile activity based on the 6 agency dimensions: goal, content, action, strategy, reflection and monitoring.

Direct instruction
The most basic form of support in formal and informal learning is direct instruction. Inquiry-based learning, especially when it takes place in out-of-the-classroom settings, is based on ill-structured inquiry activities that require a lot of support and guidance from the facilitators. This type of mobile activities facilitates such support even when participants are distributed across different locations. In the reviewed studies there is a continuum of direct instruction and guidance that work on different levels (location guidance, procedures and metacognition). These 3 levels -represented in table 1-are 3 of the 12 type of mobile activities mentioned in 4.2.
Location guidance instructions (used in 27 studies; see table 1 and Appendix A) cover types of mobile activities that explicitly move learners through different locations either in sequence or different forms of exploration. Examples of this type of mobile activities are fieldtrips or museum visits that usually are combined with technological affordances like GPS, RFID, QR codes of PDAs. Considering the dimensions of the analytical framework for learners' agency, it only provides opportunities for learners to exercise their control over the strategies (see table 2). Despite the learners are instructed about what to do (control over actions) and where to go, they can still decide how to do it. Although learners move around the place to get information, in the cases presented in the literature the information was given by the teacher or designer.
The direct instruction through procedural guidance (used in 16 studies; see table 1 and Appendix A) help the learner to carry out an activity or process. It helps learners executing and automatizing step-bystep processes. For example, in (Hsiao, Chang, Lin, & Wang, 2016;Suarez, Ternier, Kalz, & Specht, 2015) learners received question-guided instruction in order to help them during the data collection process. Other examples used list items or scripts to guide specific parts of the inquiry process King, 2007). Regarding the dimensions of agency, the learners do have more autonomy to use their content, however their control over actions, strategies and goals is limited by the step-by-step guidance. In this case also the type of mobile activities helps learners to monitor and reflect upon their learning. In case like in Suarez et al. (2015), learners could visualize their performance after the activity, so they could reflect on action (monitoring of cognition see Akyol & Garrison (2011)).
The last level of direct instruction is the metacognitive guidance (used in 4 studies; see table 1 and Appendix A). It guides the students on what cues to reflect upon and how to construct mental models. Some examples of type of mobile activities that support metacognitive instruct learners on how to carry out interpretations (Ahmed & Parsons, 2013). Other examples provide scaffolding on how to understand key scientific concepts like animal grouping (Looi et al., 2014;Parr, Jones, & Songer, 2004). This type of direct instruction allows learners to have more control over actions, strategies and content, because this main goal of this type of direct instruction is to reflect about their cognitive processes.

Access to content
The ubiquitous access to information is a common practice in our daily life, in which users seek, store, organize and categorize the information whenever they need it (Pachler, Cook, & Bachmair, 2010). In the context of Inquiry-based Learning, contextually relevant information integrated with prior knowledge (Rogers & Price, 2008) becomes essential to support ongoing observations and to create new knowledge in the physical environment. To this end, the appropriation of digital literacy is key to elicit good interpretations of the content delivered or consumed through the mobile device. In the reviewed studies, we found two types of mobile activities for accessing content (fixed and dynamic). Browsing, filtering, interpreting information found on the web, remote databases, concept maps, discussion forums, KWL 1 tables, online blogging, wikis and social networks. Table 3 shows a summary of how many times and how these type of mobile activities has been used. The fixed content type (used in 25 studies; see appendix A) refers to the type of mobile activities in which the content has been prepared before the inquiry activity starts. Although the content can be stored online, in the literature it is usually stored in the mobile device to enable also offline access. For example, through mobile functionalities like e.g. RFID tags (Chu, Hwang, Tsai, & Tseng, 2010) or QR codes (Eliasson, Knutsson, Ramberg, & Cerratto-Pargman, 2013;Nouri, Cerrato-pargman, & Zetali, 2013), learners can access relevant information for the inquiry process. Examples of activities like jigsaw (Mulholland et al., 2012), scavenger hunt games or geocaching activities (Jones, Scanlon, & Clough, 2013) integrate this type of mobile activities while keeping the inquiry nature of the activity. From the analytical framework perspective, if the content consumed has been prepared by the facilitator before the inquiry activity starts like i.e. pre-loaded e-libraries (Ahmed & Parsons, 2013) or content uploaded to a mobile app (Ternier, Klemke, Kalz, & Specht, 2012), the opportunities for learners to setup their own goals can be reduced (see table 4). This is because learners will not be able to steer the inquiry process in the direction they wish to. It has been proven that ill-structured processes require a lot of guidance (Kirschner et al., 2006;Lazonder & Harmsen, 2016), but providing guidance is not strictly the same as delivering content. Guidance can be provided on how to explore learners' objectives in an optimal way, whereas delivering specific content implies giving learners everything ready to use, which in turn, can move learners away from their own learning objectives. Thus, agency over content is also not well M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 11 supported by fixed content type of instructions. The next dimension for agency refers to the opportunities for reflection while accessing content. Providing fixed content to the learners generate less chances for learners' reflection. With fixed content the learners focus on using the information given on the inquiry process, whereas with the dynamic access learners not only need to use the information but they also need to decide how suitable is the content for the purpose of the inquiry. This does not neglect the reflective capabilities that might be enacted through fixed content, but it emphasizes the bigger opportunities for learners' reflection when using dynamic access to content.
The second subcategory is dynamic content access (used in 25 studies, see table 3 and Appendix A). It refers to the type of mobile activities in which learners actively access digital sources to browse, filter, retrieve, evaluate, and interpret content that can be valuable to advance their knowledge (Liu, 2009;Shih, 2010). Another example of dynamic access to content can be the use of remote repository of information (Huang, Lin, & Cheng, 2010;Lai, Wu, Chou, & Lai, 2010), which are not as wide as the information that can be found on internet but learners still need to browse and to filter the information. These types of mobile activities -that enable learners to manage their information (dynamic access to content)-open up the opportunities for learners to exercise their agency (see table 4 and Appendix B). First the control over the goals, it depends on the design around the use of this type of mobile activity. For instance, in Song et al. (2012), authors followed a goal-based approach in which KWL tables are essential to let the learners establish their own learning goals and plans. In this case, learners had the freedom to select their own information resources. Regarding the dimensions of content, actions, strategies and reflection, the learners have intrinsically control over all of them. The reflection dimension is covered as the learners need to developed digital literacy skills like filtering or analysis in order to find suitable information online.

Data collection
Since the appearance of the first handheld devices, data collection processes have played an essential role supporting the development of complex reasoning (Parr, Jones, Songer, & Arbor, 2002) and minimizing the complexity of collecting accurate data in the field (Parr et al., 2004). Defined as the process of digitalize, persist in time and organize the data in a digital shared space, data collection processes are essential to any scientific investigation. From table 5, one can see that the great majority of the reviewed studies (57/62) included the type of mobile activities involving data collection processes (see Appendix A for detailed information). Some examples of cooperative data collection are traditionally activities like taking notes (Hung et al., 2013;Hwang, Wu, et al., 2011;Liljeström, Enkenberg, & Pöllänen, 2013;Shih, 2010;Song, 2014;Tan et al., 2007;Yang & Lin, 2010) or drawing ideas, schemas or concepts (Chiang et al., 2014a(Chiang et al., , 2014bChung et al., 2010;Looi et al., 2014;Song, 2014;Song & Kong, 2014;Sun, Looi, Wu, & Xie, 2015). However, mobile technology also offers specific affordances like annotating data samples taken, or conceptualized mental models in order to make them shareable (Chang et al., 2012;Hung et al., 2012;Razikin et al., 2009;Sha et al., 2012;Shelley, Dasgupta, Silva, Lyons, & Moher, 2015;Song et al., 2012;Yarnall et al., 2006). From our analytical framework point of view (see table 6 and Appendix B), most of features of learners' agency are covered in the type of mobile activities involving data collection. Learners are free to decide what to capture, how to do it and after the capturing the facilitators can ask the learners to reflect and to do interpretation about it. Though, there is a difference on the opportunities for monitoring their progress between the cooperative and the collaborative data collection processes provide. Before explaining such difference, the cooperative data collection processes refer to the most common form of data collection, in which students individually capture multimedia data for later analysis or interpretation. Among the traditional techniques for data collection -taking notes, shooting videos, pictures or audios-, more specific functionalities like annotating data samples or conceptualize mental model are also widely used in inquiry-based learning. Additionally, these technological affordances have been used in specific phases of the inquiry model e.g. brainstorming or idea collection for further investigation.
Conversely, the collaborative data collection processes refer to the type of mobile activities in which students collectively associate concepts, i.e. concept maps of jointly collected graphical visualizations. Technically, the main difference with the cooperative data collection is that the data is not automatically integrated, processed and merged into a unique data instance. Cooperative data collection requires a postanalysis of the collected data, while collaborative data collection allow learners to interpret the results and react immediately to them. Examples of collaborative data collection processes are i.e. associative concept mapping (Chiang et al., 2014a(Chiang et al., , 2014bLooi et al., 2014;Song, 2014;Song & Kong, 2014;Sun et al., 2015) or jointly generate graphical visualization of the collected data during an inquiry (Rogers & Price, 2008;Sharples et al., 2015). In this particular case, the values recorded by the students were processed and combined in real time and shown to them in the for of a graph. From the learners' agency point of view (see table 6), collaborative data collection processes enable monitoring opportunities. For instance, as stated in (Rogers, 2006), learners could reflect and analyze the gathered data through collaborative data collection processes during their outdoor discoveries while still in the woodland. Such

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
13 opportunities enable a reflection in action (collaborative data collection), rather than a reflection on action (cooperative data collection) that happens when the data collection process is over (Munby, 1989).  Sharples, Taylor, & Vavoula (2010) stated that the use of mobile technology for learning situates knowledge at the center of a distributed system, in which learners hold conversations across multiple contexts. Such distribution supports the concept of shared ownership that claims that agency lies with the community of learners, rather than with the individuals.

Asynchronous 15
Forums, online discussion panels, online platforms without notifications and social boards.
Out of the 62 reviewed studies, 20 included mobile activities to steer communication between students or between teachers and students. Table 7 shows that these studies refer to a great extent to learners interacting socially to achieve certain learning goals. In Jones, Scanlon, & Clough (2013), a system that provides means of communication was used in four projects. The results provided different levels of learners control (see table 8) regarding the choice for inquiry and process (content and strategy), ranging from limited control over the inquiry topics and process -in a birdwatching activity-until almost full learners' control (besides the design of the caches) -in a geocaching activity-. In these examples, the communications were asynchronous. Learners were not notified about the communication, so it was difficult to follow up the communications. More examples of asynchronous communications can be enact with the use of forums (Chiang et al., 2014a(Chiang et al., , 2014bJones et al., 2013;Lai et al., 2010;Liljeström et al., 2013), online discussion panels (Marty et al., 2013;Wong, 2013) and comments in social boards Aristeidou et al., 2015;Boticki et al., 2015;Song, 2016;Song & Kong, 2014;Wong, 2013).
The synchronous communications rely on push & pull technology to increase learners' awareness about recent contributions in the inquiry or new messages available. These communications are volatile, elusive and they occur in real time, thus features like notifications are needed to keep the flow of the conversation. Some examples of this type of mobile activity can be found in studies that used instant messaging together with notifications (Chiang et al., 2014a, 2014b, Song, 2014, 2016Song & Kong, 2014;Suárez, Ternier, Prinsen, & Specht, 2016). Comparing synchronous and asynchronous interactions, they do not only differ on the underlying technology. Both provide opportunities to reflect, but the asynchronous communications allow learners to have more quality time for reflection. They are not urged to keep up the flow of the conversation, so they can elaborate more the message they wish to convey. This type of communication might be preferable when the goal is to achieve higher order cognitive skills (Garrison, Anderson, & Archer, 1999). Table 8. Learner's agency dimensions for types of mobile activities providing peer-to-peer interaction support Peer-to-peer interactions goals content actions strategies reflection monitor Social asynchronous Social synchronous 5.5 Contextual support De Jong et al. (2008) used the term contextual filters to refer to the support that mobile technology offers to learners filtering by the context the resources that can help them on their learning. Such contextual filtering or support, allow learners to explore, discuss and construct knowledge in real world scenarios (Donovan, Bransford, & Pellegrino, 1999). It bridges the gap between the learners and the environment that surrounds them.

Contextual support Number of studies Technological affordances
Augmented experience 12 GPS, timer or accelerometer Immersive experience 9 GPS, big display for immersive simulation, stargazing simulation, 3D weather simulation.
Adaptive feedback 7 intelligent systems that give hint for making further observations Only 16 of the 62 studies included mobile activities for contextual support (see Appendix A). Table 9 shows the three subcategories (augmented experience, immersive experience and adaptive feedback). First, the augmented experiences (used in 12 studies, see table 9) prompt information to the learner when he/she steps into a specific location or when certain amount of time has passed. Thus, the context, through this type of mobile activity, creates the learning conditions for students, but the student has the control and decides what to do or to explore (Nathan & Robinson, 2001). The majority of the designs used the GPS to model this type of activity (Boticki et al., 2015;Chiang et al., 2014a, M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 15 2014b; Dunleavy et al., 2009;Huang et al., 2010;Hwang, Chu, et al., 2011;Jong & Tsai, 2016;Kamarainen et al., 2013;Laru et al., 2012;Schaal et al., 2012;Wong, 2013) (see table 10 and Appendix   B). This type of mobile activities does not offer opportunities to monitor the learning progress, because they focus on the delivery of content through the context. Therefore, Second, the immersive experiences (used in 9 studies) refer to the type of mobile activities embedded in a digital simulation. Digital simulations are frequently used in inquiry-based learning to explain complex concepts or processes that cannot be easily reproduced in real life. Examples of this type of immersive experiences have been reported in (Chen & Lin, 2016) in which students could engage in a free-form of stargazing with simulated constellations or in (Lui, Kuhn, Acosta, Quintana, & Slotta, 2014) in which students collected data in rainforest immersive simulation. Another example is the 3D weather simulation (Hsiao et al., 2016), in which learners combine factors like temperature or pressure to learn the implications those have in the weather. All in all, these immersive experiences are characterized by having learners interacting with a digital environment that augments their learning experience. Concerning learners' agency, these types of mobile activities showed options for learners to control all the agency dimensions.
Last, the adaptive feedback experiences (used in 7 studies) offer incremental support based on learners' performance. In this case, learners' agency is supported but some considerations must be made depending on the inquiry activity. In (Chu, Hwang, Tsai, & Tseng, 2010) learners were guided to do field observations in conditional steps. Based on their performance they received different hints to continue the exploration. During the activity, the observation and the hints received were recorded so learners can monitor their steps. In this example, the learners' agency is limited because they received information about strategies to explore the field. However, in (Hung, Hwang, Lin, 2013) they used an automated feedback meant to motivate learners' reflections. In this case, no constraints were found for goal, content or strategy decision on the inquiry activity. In both cases the control over the actions might be influenced the adaptive feedback.

Discussion
The inclusion of mobile technology to support inquiry-based learning processes, has provided ubiquitous opportunities for learners to carry out inquiry processes in a more self-directed way. However, little is known about the degree in which the type of mobile activities in inquiry-based learning, provide real opportunities for learners to exercise their agency. Additionally, although research pointed out the need for tighter guidance and scaffolding in inquiry-based learning and open ended processes (Alfieri et al., 2011;Hmelo-Silver, Duyncan, & Chinn, 2007;Kirschner et al., 2006;Lazonder & Harmsen, 2016), there M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
are not so many practical strategies for practitioners to enact the right balance between learner's agency and scaffolding. Thus, this review provides a strategy to identify the level of agency support for a given type of mobile activity.
In order to deepen our research in the learners' agency (Bandura, 2001;Zimmerman, 1990) -an aspect considered to be essential for inquiry-based learning-, our analysis implemented a three-step process analysis. It started with a state-of-art analysis of the ubiquitous properties in existent mobile frameworks. It revealed that in the context of inquiry-based learning 12 types of mobile activities for inquiry-based learning can be found (see Appendix A). In a second step, to characterize learners' agency 6 dimensions were defined. The dimensions consisted of: 1) learners' control over the goals, 2) learners' control over the content, 3) learners' control over the actions, 4) learners' control over the strategies, 5) learners' opportunities for reflection and 6) learners' opportunities for monitoring their own learning progress. The last step is the actual analysis of the 12 types of mobile activities using the 6 agency dimensions. In Appendix B, while the types of mobile activities that support more learners' agency are placed in the positive side, the ones in the negative side offer less opportunities for learners' agency (or more scaffolding for learners).

M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT a given agency dimension. Conversely, a cross means that there has been an evidence that a mobile activity limits or constraints the control of learners over a given agency dimension. This second case must be interpreted with caution, because for some types of mobile activities a cross can have a positive interpretation with regards the balancing of support of agency and scaffolding. One example is the agency dimension named content in relation to the fixed content type of mobile activity. This example shows an obvious limitation of learners' agency while accessing fixed content, because the learner has no control over the content used for the inquiry as it is given by the teacher through the mobile technology. However, although it limits agency, it is a great opportunity for teachers to provide content scaffolding in order to ensure that learners work with desirable concepts. This interpretation can be generalized for the goals, actions and strategies dimensions too, because in case that the agency is limited or constrained it provides opportunities for guiding learners on their learning. For example, with mobile activities that support metacognitive guidance learners will have certain degree of freedom, but usually the mandatory reflection moments guide or lead learners to specific learning goals. Similarly, the mobile activities that support augmented experiences are suitable for learners to explore the environment, making sure that they stumble upon some desirable concepts needed for their inquiry (often placed in a location or triggered to the learner at a given time). Another example, that shows how the analytical framework can be used to better combined scaffolding with learners' agency can be seen in the adaptive feedback. The learners can have control over the inquiry activity but still will be asked to perform specific actions based on their previous performance. This, again, is another type of guidance through mobile technology.
A different interpretation needs to be done for the reflection and monitoring agency dimensions. The meaning of the information in table 11 still represents is the types of mobile activities enabling monitoring or reflection. But, unlike the other agency dimensions, a cross in monitoring or reflection means that there is no technical support to monitor or to reflect about their progress. In this sense, techniques like e.g. Learning Analytics can provide means through their dashboards to enhance the support of theses agency dimensions. The learners can trace and visualize their progress over time (Schwendimann et al., 2016) which allows them to monitor their progress. Such monitoring implies more control for learners over their learning to reflect and also to adapt their behavior based on an informed decision.
The analysis of the learners' agency support through mobile activities can have implications for educational practitioners. Through the analytical framework, they have more means to strategically design their mobile activities based on the agency support. It is possible to aim at specific dimensions of agency and to provide guidance for other dimensions in which learners have a deficit on. For example, learners with problems to focus on their learning goals should be supported with mobile activities that provide guidance on pursuing learning goals like e.g. direct instruction or fixed access to content. That is to say, those learners should not use mobile activities in which they have control over their learning goals like e.g. data collection or mobile activities that offer contextual support. With this in mind, it is up to the teacher to use the analytical framework and design activities that, first provide direct guidance and then later gradually design new ones e.g. based on contextual support of peer-to-peer interaction, in which they have a higher degree of agency.
Taken together, the findings presented in this study, the analytical framework and the examples of M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
how to use it, contribute to overcome the criticism about discovery learning processes or methodologies that enable more learners' agency (Kirschner et al., 2006;Mayer, 2004). Because it presents practical instruments for practitioners to optimally balance learners' scaffolding and agency through mobile technology, without jeopardizing the natural capacity of learners' to be autonomous on their learning (E. Tan, So, & Zhang, 2012). These findings also align with the claims made in a recent meta-analysis on inquiry-based learning, that pointed out the relevance of providing instruction and support during the inquiry-based process (Alfieri et al., 2011;Lazonder & Harmsen, 2016). Finally, this analysis provides two main outcomes in relation to the state-of-art of mobile inquirybased learning and its implications supporting learners' agency. First, it provides a summary of the most typical types of mobile activities in inquiry-based learning, as well as the technological affordances used to enact them. The second outcomes of this review is the analytical framework that can be used by practitioners to design more balanced mobile activities in inquiry-based learning. This instrument allows them to design mobile inquiry activities that gradually foster the different dimensions of learners' agency. Moreover, it helps to raise awareness about the potential of mobile technology to enable learners' agency. In line with Vogel's criticism (Vogel et al., 2009), the inclusion of mobile technology on inquiry-based learning is not enough to have autonomous and seamless learning processes. It requires agentic learners that make a step further and take control over their learning. To this end, our analytical framework and the analysis done contributed to identify and promote the inclusion and usage of suitable types of mobile activities that support learners' agency.

Conclusions
This review aims at examining to what extend the use of mobile technology for inquiry-based learning supports and limits learners' agency. To this end, 62 studies about mobile inquiry-based learning have been analyzed through the types of mobile activities included in their designs. In a second step of our evaluation, the analytical framework extracted from the definitions of agency in the literature revealed 6 dimensions for learners' agency (control over goals, over content, over actions, over strategies, and options for reflection and monitoring). These were used to evaluate 12 types of mobile activities derived from the direct instruction, access to content, data collection, peer-to-peer interaction and contextual support type of mobile activities. As a result, the review provides an indication of what aspects of the learners' agency can be exercise with the different types of mobile activities (see Appendix B). Despite the exploratory nature, the contribution of this review is twofold: first it contributes a systematic review of the types of mobile activities used in inquiry-based learning (see Appendix A). Second, it provides a preliminary analytical framework (see section 3) that can raise awareness about the role of mobile technology to support learners' agency and scaffolding.
There are several limitations associated to this study. The majority of the studies found were case studies because of the lack of a larger body of experimental research studies. This has made it difficult to draw conclusion about the effectiveness of these types of mobile activities supporting learners' agency. Another limitation comes from number of articles analyzed, which was not exhaustive. For example in our search we excluded the term discovery learning which was merged with inquiry-based learning in 1998 (De Jong & Joolingen, 1998). Including this term in our research might be resulted in a more M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT comprehensive list of analyzed work. Also the reliability of the analytical framework has not been discussed thoroughly and it suggests a follow up investigation to validate the agency dimensions. Additionally, the age range of the students and the relation between learners' agency and scaffolding (without mobile technology) have not been taken into consideration. Nevertheless, these limitations do not alter the main outcome of the study.
Recent meta-analysis on inquiry-based learning (Alfieri et al., 2011;Lazonder & Harmsen, 2016) have pointed out the importance of instruction and support in the inquiry process. The use of mobile devices for inquiry-based learning is able to foster different types of mobile activities combining a mix of instruction and agency for the learners. This review has distillated suggestions for such combination to be successful. For example, the use of metacognitive guidance, as part of the direct instruction mobile activities, provides a good balance between learners' agency and learners' scaffolding. The dynamic access to content, as part of the type of mobile activities to access content, offers some chances for learners' agency, but also helps to improve learners' digital literacy skills. The type of mobile activities for contextual support provide some opportunities for learners to control aspects of their learning. Nevertheless, there are some types of contextual support that can undermine the learners' control over the content and their actions. The last two types of mobile activities, data collection and peer-to-peer interaction, allow learners to have most of the control over their inquiry process. However, those types of mobile activities in order to be successful require a good combination and balance with the other three types of mobile activities: direct instruction, access to content and contextual support.
Taken together, this review provides a preliminary work to better understand the influence of mobile technology on learners' agency for inquiry-based learning. Several reflections emerge from this study advocating caution while finding the optimal support and balance between learners' agency and scaffolding. First, we advise for more teachers' and learners' reflection about the use and appropriation (Pachler, Cook, & Bachmair, 2010) of mobile technology for learning. Such reflection can improve the skills of self-identifying learning strategies and consequently lead to more genuine agentic learning (Song et al., 2012). Second, more experimental research is needed to find the optimal proxy for learners' agency, not only for mobile inquiry-based learning but also in broader contexts.
Finally, in the fast-growing field of wearable applications and sensor devices for real world data collection and activity tracking, we foresee further possibilities for intelligent support for inquiry-based processes. This will call for new analysis of learners' agency, in which the study of learners' agency done in this review could be taken as a basis for further improvements.