Semantic level instrumentation JavaScript Client Library enables realtime continuous streaming of all user’s interaction process data that has technology that automatically caches data for spotty networks (such as schools) to ensure data integrity.

Since this atomic level data is timestamped, it is granular enough to be fed to machine learning models to visualize behavior at scale, playback through data, and generate data science rooted inferences about activity that indicate both cognitive and non-cognitive path patterns (affect) that understand not only how a user is thinking and what their competencies are but also how they are feeling (boredom, frustration, confusion, etc.). Some uses for these capabilities include adoption, social-emotional learning, collaboration, and complex problem solving assessment.

https://developer.metacog.com/logger/overview
https://developer.metacog.com/logger/instrumentationguide

DataRequest API allows users to retrieve the original event streams as they were logged by the widget for a set of filtered sessions.The validated events are stored in the main database, without any restrictions on the amount or age of the records. Once all the data is stored in the main database, it can be retrieved from, and processed by different services: from raw data extraction to complex Educational Data Mining processes and visualizations tools, to API's, metacog offers you a wide range of options for extracting meaningful information from your data.

The retrieval-data API is a JSON-based REST API that allows users of the metacog platform to request data originated in learning objects instrumented using the instrumentation input API.

Given the large amount of data logged by each learning object, the results of a query are not returned immediately. Instead, a data-request object is created in the server to keep track of the current request, based on the filter parameters defined by the client, while the real processing is executed as a background process.

The API allows polling the data-request object in order to know the status of the background process. The client may also wait for an automatic e-mail notification to a prevalidated e-mail address, indicating that the process is finished, and the data can then be downloaded.

This tutorial presents the VisualizationRequest API as one of the ways to retrieve information stored in the Metacog Platform by instrumented widgets, and how the Client Side Library offer helper methods to interact with the API to build aggregated reports via Javascript.

The intended audience are both the Domain Experts who want to understand what kind of aggregated datasets are available out-of-the-box and the Front End Developers who need to build visually-appealing reports based on those datasets.

Once you have an instrumented widget and your learners start to use it, an amazingly huge amount of rich data will begin to being stored in the Metacog Backend.

Currently, the Metacog Platform offers two ways to retrieve information, with more services being under active development. The VisualizationRequest API allows you to retrieve aggregated datasets with different views of your data. They are useful for general data-exploration without the need to download and process the original event streams.

Metacog Training Toolbar is used to record and playback Training Sessions, on an already instrumented widget, and in the context of a particular Learning Task. This service provides the ability to score open ended performances via rubrics.

Target Audience is learning experts who want to create Training Sessions for Learning Tasks, in order to train the system with the goal of producing automated real-time scores from Learner Sessions. A Score is an object that represents an association between a Rubric and a Training Session, and holds specific values for each dimension in the Rubric.

In scoring mode, the goal of the learning expert is to visually check the event stream of a Training Session and based on what he sees, make decisions about the values for each dimension. Also, he/she can mark the event stream with special markers (indicators) to enrich the scoring information.

Sending scoring results in a format available by recommendation engines such as Knewton, etc.

Since recommendation engines are built to probabilistically adapt to where the user is in their learning progression for a given subject matter (a directed acyclic knowledge graph) then feed content or questions to their zone of proximal development, they have been limited in the past to multiple choice questions due to complexity. Metacog enables more authentic and real-world measures of competency being assessed in digital environments to be utilized to add a much deeper and wider range of learning objects capable of providing individualized responses. This allows product differentiation through enabling much more engaging content to be utilized to improve adoption and churn rates, as well as simultaneously testing complex skills.

The ability to markup user sessions to train a machine learning model to automatically detect behavior patterns such as boredom, frustration, WTF and engagement.

Since this atomic level data is timestamped, it is granular enough to be fed to machine learning models to visualize behavior at scale, playback through data, and generate data science rooted inferences about activity that indicate both cognitive and non-cognitive path patterns (affect) that understand not only how a user is thinking and what their competencies are but also how they are feeling. Some use cases for these capabilities include improving adoption, social-emotional learning, collaboration, and complex problem solving assessment.

https://developer.metacog.com/logger/overview
https://developer.metacog.com/logger/instrumentationguide

Playback through data streams.

Since this atomic level data is timestamped, it is granular enough to be fed to play back (as a data fed “video”) when proper instrumentation is implemented. Sessions are captured in data and can be played back on demand to support how scores were generated and to prove micro-competencies/badges were earned. These sessions are also used for scoring training to teach machine learning algorithms what constitutes desirable rubric framed behaviors and scored utilizing the scoring harness mechanism.