Skitag AI Ethics

Data Governance and Privacy

The user is the owner of the data. The data is stored locally on-device. The data is synced within the user's devices connected to its iCloud ID account. Behavioural data could be classified as biometric data. Nevertheless, the data stored do not contain any information that could be used to identify the user. The user can delete it anytime.

Explicability and Transparency

Whenever it is possible, Skitag AI technology will be publicly available. The user can have access to graphical representation of their data. The user can watch and understand the meaning of the input data. The user is responsible for the labelling task of the training data. That means that the user has some understanding of the meaning of the data and what the user is training an algorithm for. Training metrics will be openly shared to the user. The user can watch/evaluate the performance of its algorithm with its own evaluation data.

Skitag AI - LSTM Github repository

Bias and Fairness

The user will collect and train its own AI engines. The information contained in the users' datasets will be explained (i.e. on-device sensors data collection and interpretation) as well as the algorithm life cycle (i.e. data collection, training, evaluation, deployment and interpretation of their outputs). The AI engine will be biased showing the users' own carving biometrics. The user will be the owner of its own AI algorithm. Whenever it is possible, the user can choose to publicly share the resulting algorithm with others, so it can become an alternative AI algorithm baseline that other users can choose to use (i.e. The user can use it, share it, or both).

AI System lifecycle

What is "The Data"

Skitag uses the iPhone or the Apple Watch sensors to collect behavioral data (i.e. GPS, Accelerometer and Gyroscope). For instante, Skitag uses Location updates (e.g. speed, altitude or distance) and movement measures (e.g. attitude, accelerometer or gyroscope data) to decide whether the user is going downhill or not, or is carving or not. Moreover, Skitag collects GPS location information (e.g. latitude and longitude). This information is used to locate the user in a map, and to draw its path while skiing (i.e. trace a downhill and plot into a map).

Who owns the Data?

The user is the owner of the data. Skitag AI does not have access to the Data.

Where is the Data stored

The Data is stored locally on the device used to collect the Data; i.e. an iPhone, or an Apple Watch. Then, the Data is synced among the other devices associated with the user Apple's iCloud ID.

Why do I need to store the Data?

There are two kinds of Data within the Skitag ecosystem. First, there is the Data related to a Ski Session; i.e. Distance, Speed, Efficiency, Balance, Turns, etc. This information is stored to let the user know their performance, and to have a history about their behavior as a skier.Then, there is the Data related to the user behavior; i.e. Location and Movement. The storage of this Data is optional. The user can decide to store or not that Data. In case the user decides not to store the Data, the Skitag App does not record that information. Otherwise, if the user accepts to collect that Data, the Skitag App stores it locally and then syncs it through iCloud (private container) to the other connected devices.

Why should I store the Behavioral Data (biometrics)?

You should want to store that information in case you would like to fine tune the machine learning algorithms measuring your carving characteristics. By doing this, you’ll have a customized algorithm that works nearly perfectly with your own behavioral Data (i.e. learns to evaluate your own biometrics).

Does anybody else have access to the Behavioral Data (biometrics)?

No, unless you decide to share it. In that case, you’ll be sharing that information with known users with a known Apple ID.

What is a CARBIO?

It’s a machine learning algorithm that receives an input (e.g. speed, altitude, acceleration, attitude, etc.), process that Data, and outputs a binary decision (e.g. carving, not carving). The machine learning algorithm is a Long Short Term Memory (LSTM) Recurrent Neural Network architecture. It is used to evaluate the kind of information that has a timeline dependency.

Does the CARBIO have the same status as the Data?

Yes. The user is the owner of the CARBIO. It's stored locally on the device and synced by iCloud to the other devices SIGNED IN to the same Apple's iCloud ID. Finally, the user can share the CARBIO to other users so they can try their biometric (i.e. their Behavioral Data) with that CARBIO. In case the other user fine-tunes the CARBIO with its own Data, then the new CARBIO is now of its own; i.e. the CARBIO has changed, and now contains information from the user who fine-tuned the CARBIO.

What kind of tasks can I do in any of the synced devices?

iPhone: Collect, process and output information about your ski session. It uses the Location and IMU (Inertial Movement Unit) sensor to evaluate your behavioral Data (biometrics) and outputs information about your downhill, and about your ski session. You can watch your biometric Data, you can label your Data, and use it to fine tune (or train) a machine learning algorithm (a CARBIO). You can sync that information to your connected devices (e.g. iPad). iPad: You can watch your SKI SESSIONS and DOWNHILL data. You can watch and label your behavioral Data (biometrics). You can fine-tune (or train) a machine learning algorithm (a CARBIO). You can sync that information to your connected devices (e.g. iPhone).