Literature Review
2.5 The ‘learning-contexts’-based adaptive m-learning generation
The main difference between this generation and the previous two is that applications within this generation take into account users’ learning contexts for determining which learning materials or activities students should be given to perform. The values of these learning contexts are requested by the application to be inputted by the user, i.e. an interactive method of retrieving contexts.
My mCALS suggestion mechanism framework is related to the foundations of these systems in this generation. This is because, like the other applications in this generation, there is common aim to suggest appropriate learning materials to students based on their situation. However, the suggestion mechanisms in this generation are not context-aware, and I wish to construct my framework so that it has context-aware capabilities in order to increase benefits for learners, such as to minimize the need for students to provide input to the mobile device whilst ‘on the move’.
However, if the learning schedule proactive approach of my framework were to be unsuccessful, then my framework would be ‘learning-context’-based (i.e. this generation), rather than ‘learning-contexts’-aware (i.e. the next ‘learning contexts’- aware adaptive generation). ‘Learning-contexts’-aware adaptive applications are described in 2.6. Note that context-based applications only require an additional feature in order to be developed as context-aware. However, authors/developers may want to develop context-based applications instead of context-aware applications due to various reasons, some of which are discussed below.
Three main applications/research works included in this generation include TenseITS (Cui and Bull, 2005), CoMoLE (Martin and Carro, 2009) and didactic profiling (Becking et al., 2004); these are described below along with other
Cui and Bull’s (2005) TenseITS application
This application focuses on providing English learning materials for Chinese students to learn in their available time. Four learning contexts are taken into consideration – location, available time,concentration level (at the beginning of the session) and the
frequency of interruption (at that location). The learner’s user model is also
considered when learning materials are selected for students. The attributes of the user model include their knowledge level, misconceptions of the English language and
difficulties in learning the language. The attributes of the user model are constructed
continuously from the user during their interactions with the application. A similar system prototype (Bomsdorf, 2005) also selects appropriate materials for students based on the four learning contexts, a slight difference being the frequency of interruption replaced by frequency of disruption.
Cui and Bull (2005) pointed out two reasons why they employed an interactive multiple choice method, rather than deploying a proactive method by means of, for example, retrieval from the student’s electronic diary. The first reason was that students often did not conform to their schedule as observed by their absence from lectures, so the information retrieved from their electronic diary may not be accurate if it was used for obtaining their available time and location information at a specific point in time. The second reason was that the authors’ system was designed for use within short periods of time and primarily in-between other activities which students may not have recorded in their electronic diary, even if they had kept one. Therefore, the authors noted that the location may not be detected accurately because this was not recorded. Similarly, there was no way of inferring the learner’s available time.
The application operates by first requesting the user to input the values of the four learning contexts, which are to be selected from multiple choice answers, before each learning session. A set of suggestion rules are built-in to the application to determine which learning materials are appropriate to learners based on the context values and their user model. Subsequently, learning materials are recommended to users when they wish to learn/study. This set of suggestion rules is described in 4.4, where I discuss the recommendation of learning materials appropriate for different contexts.
The future work of the authors includes extending their system “to other areas of English that Chinese students find difficult, for example: the use of articles” (Ibid), as well as for Russian or Arabic speakers, as these students may also have difficulties with tenses and articles, or other languages. Their system is particularly good for “[a]ny language or aspect of language that can be tested with multiple choice questions (because input on a handheld device is difficult), and where students commonly have difficulties, could be potentially useful” (Ibid).
The TenseITS prototype has not been evaluated and the authors noted that “the feasibility of extending the system in different areas and for different target groups, needs to [also] be tested” (Ibid). I contacted the author of this work to determine whether we could discuss this research further; however, the authors declined as they are no longer continuing with this work. No further work relating to this topic has been published since Cui and Bull (2005).
The CoMoLE suggestion mechanism has been designed for recommending appropriate learning activities to learners where the recommendation process is dependent on both the user’s internal and external learning contexts. The user’s internal contexts include the learner’s profile (such as their LS, preferences and previous actions/interactions with the application). The user’s external contexts include their location, available time and mobile devices used as well as devices available to them. It also takes into account the fact that users may use different physical devices (such as PCs, laptops, mobile phones and PDAs) and thus activities are adapted appropriately to the different device types. There is an option which, if appropriate, according to the user’s learning contexts, would interrupt them and alert them to the availability of an activity. The system also allows collaborative activities between users to be performed. The system could accommodate both individual and collaborative learners. If the learner is conducting collaborative learning, then their partners’ internal and external contexts are taken into consideration for the selection of appropriate materials.
A number of courses have been incorporated into the CoMoLE environment: A ‘boolean algebra’ course, which was described in Martin et al. (2007),
describes how individual and collaborative activities are adapted or suggested to users based on the users’ learning contexts and preferences. The types of activities include theoretical examples, interactive examples (simulations), individual tests and collaborative activities.
Two subjects, “data structures” and “operating systems”, were described in Martin and Carro (2009). These were used by students to learn/study with and also formed their two evaluative case studies. Different types of learning activities related to these subjects were included. Students could use different
devices such as PCs, laptops or PDAs to access and perform these learning activities. Results of the case studies are discussed together with the results of our Java LOs validation study in chapter 7.
Note: Most of the suggestion mechanisms need to be (at least in part) content- specific, as they have to be teaching something. These materials can be adapted or changeable in the system. This requirement normally exists to ensure that the quality of the content is sufficient, and an expert teacher is normally required to check the quality of the content.
Becking et al. (2004)’s didactic profiling framework
This didactic profiling framework is a generic standardized mechanism which can be deployed by researchers/developers. It defines a set of contexts that should be considered for determining the types of learning materials/activities for learners in different situations. It is centred on an inference engine and contains a set of filtering rules, which are based on learner profiles and the characterization of LOs. The learning contexts used within this mechanism are classified into the following four categories - situation, learner,LOs, and participation. However, exact details of the
filtering rules for their inference engine were not presented.
1. The situation category contains frequency of interference (during a learning
session), available time (scheduled or estimated), equipment at disposal
(learning tools, aids, books, other learning materials which can be used in the situation) andrestriction of action and expression (for example, restriction to
read, write, listen or speak in that situation). The first two learning contexts were deployed in Cui and Bull’s (2005) application.
2. The learner category includes level of concentration/distraction (self-
evaluated ability to keep concentration despite environmental interferences), previous knowledge relating to topic, and previous knowledge relating to
technology. The first two learning contexts were also deployed in Cui and
Bull’s (2005) application.
3. The LOs category includes instructional goals (standards appropriate for the
conditions of mobile learning) andlearning content.
4. The participation (also known as collaboration with peers) category includes individual learning session(self-paced or supported by tutor),partner session
(working in groups of two students),group session (working in groups – self-
organised or by teacher, informal or formal).
I also contacted these authors regarding the evaluation of their framework/system prototype; however, no replies were received.
Other miscellaneous frameworks/systems
A system was developed by Cheverstet al. (2000) for tourists visiting the City of Lancaster, England, which took into consideration environmental contexts (such as the opening times of the city’s attractions and the current time of day), which were relevant for creating a tailored tour and navigating a visitor around the city. The visitor’s personal context information was also stored and used for adapting the visiting materials including the visitor’s current location, personal profile (interests, preferred reading language, set of attractions already visited) and learning style (whether active/passive role).
A situation-aware framework/mechanism has been developed by Bouzeghoub et al. (2007) that takes into consideration time, place, user knowledge, user
activity, user environment and device capacity for adaptation of learning resources to the user.