Human Digital Twin (HDT)

The human-specific variant of the Information Digital Twin 


A User-Dedicated Smart App

The Human Digital Twin (HDT) is an AI-based application that learns its user's behavior and preferences across multiple, complex interactions with their environment. The HDT thus enables users to optimize and automate their digital interactions based on their preferences and interests, allowing them to lead balanced and healthy social and professional lives.

The HDT is a kind of a "meta-app" which observes how its user interacts with digital applications and digital systems, the context of these interactions, and learns the user preferences and interaction habits. Given specific user interests and objectives, the HDT calculates and assigns an information value to various user choices. The HDT then relies on this information value to calculate how user actions might affect their objectives. It then recommends a course of action that maintains user interests while preserving their balance and well-being.

A Life-Long Companion

The Human Digital Twin is intended to assist and learn from its user throughout most of their digital encounters or scenarios. The more scenarios the HDT manages, the more accurate are its recommendations and the more insights the user has about their own behaviors, choices, and the implications of those choices.

The HDT offers its user a personalized viewpoint on their digital interactions and activities. Therefore, a user can rely on their HDT to provide them with individualized recommendations which takes into account their preferences and long-term interests.  

In addition, when connected to a Social AI Platform, the HDT can further align its user with social objectives and increase their social resilience.


The HDT Technology

The HDT's core techniques are based on reinforcement learning and Bayesian inference (RL). In summary, the HDT employs Bayesian inference to user interaction data to define the parameters necessary for an RL agent to evaluate and recommend actions. These parameters include the states relevant to an objective, their degree of relevance to the objective, the relevant actions and their value, and the action rewards.


However, the defining characteristic of the HDT is the dynamic nature of the RL environment, in which states, states' values, actions, and rewards are continuously updated to reflect changes in the user's environment, and user objectives and preferences


In addition, more advanced versions of the HDT rely on unsupervised learning to identify user preferences and decisions patterns, which improves the accuracy of the HDT recommendations.    

Digital Home 3.png

Automating Digital Choices

Once users start to follow the HDT recommendations, the HDT can begin to handle many of the user's digital decisions, thus giving the users more time and attention to dedicate to creative and social activities.

The HDT has the same features as the more general agent, the Information Digital Twin (IDT). As such, the HDT can also connect to centrally managed scenarios (e.g., health, traffic, pandemics management, or operational risk management). This connectivity enables the HDT to update its user "action information values" and thus improve the effectiveness of the user actions and increase their alignment with social objectives when necessary.

How to Start?

The HDT is implement according to user scenarios. Each user scenario is specific to the user and captures the interaction information relevant to the user. Each scenario and its features is structured in the form of an Information Heat Map (IHM) which indicates the information values-in bits-of the various interactions taking place in the scenario. The HDT can then learn the user preferences and use the IHM to determine best options for actions. 

The user can then keep joining/initiating other scenarios, which translates into expanding the IHM to include new inputs, actions and output features. 

The HDT's unique approach to data architecture enables the seamless, effective, and almost endless integration of any kind of data into the user information heat map. 

HDT Scenarios.png

Current Human Digital Twin (HDT) Use Cases 

At this point of the HDT's development, we have defined the algorithms underlying scenario representations, Bayesian inference, and the RL agent environment definition.

Based on the developed algorithms, we are maturing the HDT for two use cases:

  • Patients-HDT, which assists neonatology ICU patients,

  • Children-HDT which assures children's safety and minimizes their risk of abuse.

Both use cases focus on children because we believe modeling and forecasting risks for children are more achievable due to children less complex nature and behavior.