Update: Approving Our Research

I am submitting an IRB application as the principal investigator of our research project. The IRB is in charge of monitoring behavioral research involving humans; this includes software evaluation. We are avoiding collection of personal information by assigning students anonymous research IDs and we will not restrict our research to adult NMT students. Only the instructors will know the true identities of our participants. We have filled out the IRB Application and Form A and await approval before our tentative start date of March 20th

-- Hugo

Update: Using the System

This blog post summarizes the usage of our attendance tracking system. We have been testing it in order to make it usable for the study, and to keep information safe from leaks.

- it must be usable
- it must keep information safe from any unintended accesses (accidental or penetration, some subversions)

This should make students more willing to participate. The basic procedure for students is:

- download from Play Store (we'll support both x86 and ARM devices)
- or download to their laptop (which can run Linux, Windows, or OSX)
- on initial startup, students associate a password with their anonymous research ID
- students will use this password-ID pair to login on every startup
- students may click the bottom-most button until they reach the attendance verification screen
- students may review their list of classes and corresponding attendance records
- they finish by gathering and sending attendance evidence

The procedure for instructors is:

- download QR or Pattern image
- or copy pin or password
- display this code in a prominent location
- wait for students to register their attendance
- download attendance records as a CSV file from the server

We'll refine this process before we begin our experiment.

-- Hugo

Update: QR Decoding and External Libraries

QR codes are an important feature to the project. I have implemented image loading and QR decoding. Part of the implementation involved familiarizing myself with Qt's build process in order to make use of 3rd party libraries. Qt's image selector is specialized on iOS so that users can select a picture from the native gallery. I used the Quicknative library to provide similar functionality on Android. It uses Qt's Java interface to create Java objects and make calls to the Android API in order to provide users access to a native Android gallery for their image selection. I also used the quirc QR decoding library, which searches a given image for QR patterns using a standard floodfill-based algorithm, then it uses the corners of the QR code to straighten the image with a perspective transform. I had to fork this library in order to add C++11 support and a qmake build file `quirc.pri` so that any Qt project can make use of quirc.  After importing these third party libraries, I parsed the `file:///` path, loaded it as QImage, performed some image processing, converted the image to a character array in order to  run quirc's detection and decoding algorithm. Quirc can process pictures which unambiguous represent a QR code and are  distorted only by a perspective transformation (for example, if the picture was taken from the corner of a room). In case of decoding failure, we are able to send the image to the server.  The next step is to perform QR decoding a separate thread, display a loading bar, and let users crop images before processing.


-- Hugo

Update (1/15): Client UI --- Usability Improvements

We need a usable app that works on many different screens. Now that we have designed the system, I can start refining the UI. The first and easiest way I made the cross-platform UI is by using Qt's Quick Controls. These are prebuilt UI elements that can resize to satisfy simple constraints:  for example, a ListView can be used to show a list of similar elements, and a GridLayout can arrange arbitrary Qt elements in a layout with a certain number of rows and columns.  Another part of the solution involved implementing dynamic margins and element sizes. The application has very small visual margins on smaller screens, but it will not spread out the cell-phone centric UI if it is run on a larger screen. I also implemented a useful features for small screens, the UI will have a minimum screen size after which it will display scrollbars in order to keep all content available. The Qt documentation was useful, I have also been testing and modifying the UI in order to make it easier for our users to access all of the app's features.  

Future goals include using a wider range of testing devices and insuring that all new UI elements work on many different devices.

-- Hugo

Update (1/8): Object Model Updates, SQL+JSON+Java Object+QObject, Storage

An important part of any software project involves translating information from a real-world domain to digital information in the artificial system. We have asked ourselves important questions, including: what do we need to represent? how much information do we need to represent students, attendance logs, and classes? In the Onion architecture, this information belongs in the object model I worked on developing the object model with Nico. Here is a draft of our work (objects are indicated using `*`, each object has a list of property names and types):

* course
- course id (short String like 'CSE 101', unique)
- course name (long String)
- instructor name (long String)
- location (long String)
- readable time (long String)

* users
- research id (long String, unique)
- password (long String)

* student-course junction table

* attendance log
- log id (int, unique)
- received time (timestamp)
- sent time (timestamp)
- auth key (long String)
- auth type (short String)
- auth status (ok, late, absent)
- device id (long String)

foreign keys
- attendance id (int, from the attendance session table)
- research ID (long String, from the users table)

* attendance session
- attendance id (int, unique)
- authtype (short String)
- authkey (long String)
- course id (short String)
- start time (timestamp)
- end time (timestamp)

Now we need to represent the model in code.  The data on the server must be stored in a SQL database, converted to Java Object using `sormula`, and converted to JSON using Google's `gson` library so that it may be sent from server to client Once this data arrives at the client, it should be translated to a QJsonDocument and then  into a QObject. This QObject can be easily consumed by the GUI, which is written in the QML language.

An important future goal is extensibility. We will make frequent changes to our data model as the experiment progresses. We need to translate our model to repetitive code across several platforms while maintaining consistent names and relationships. Ideally, we would have one source-code for both the client and the server. Instead, we're using JSON objects, Qt objects (defined using C++), and Java objects to store and transmit information. To help keep the data objects consistent across platforms, we are working on a script that generates (1) a SQL create script, (2)Qt Object and JSON conversion code in C++, and (1) Java object declarations in order to keep names and types consistent and easy to update.

Update 02/19/17

Work Accomplished:
We have completed the IRB forms required for our application.

Goal:
Test the student application on iOS and complete the Protecting Human Subject Research Participants course, which is required by federal law for projects involving human subjects research.


Outcome:
I learned the requirements for getting the Institutional Review Board's approval for research involving human subjects.

--Eugene

Update 2/12/17

Work accomplished:
We have shortened the proceedings file and submitted the final version of our paper. After submitting the paper, we managed to complete the presentation for the SITE 2017 conference.

Goal:
We want to complete the necessary IRB forms and related documents in order to submit a proposal for the user study we will conduct.

Outcome:
I have gained an understanding of what professional research is, and why collaboration skills are important in projects like this.


--Eugene