Helvia.ai joined the LAW-GAME consortium as an expert on digital media production that specializes in Artificial Intelligence (AI) and automation, with in-depth experience in the development of intelligent chatbots and AI for interactive storytelling. Our LAW-GAME role has been to lead the development of interactive chatbots that support the use cases of the project, i.e. that undertake the role of a subject that gets interviewed by the player.
Helvia.ai has developed the Interactive Storytelling Engine (ISE) module that controls all dialogue-based interactions with non-player characters (NPCs) in the different game scenarios and determines how the game scenario evolves based on the player’s actions. ISE is designed to accept inputs from the player and the game environment and return the effects of these inputs, e.g., NPCs’ responses and emotions, to the game engine. ISE transforms inputs to outputs based on a specific story that describes the conversation scenario. Moreover, ISE utilises “interactive” storylines instead of “linear” ones, where the scenario has the potential to unfold towards different directions depending on player’s decisions. Helvia.ai developed also the AI Narrator (AIN) module that acts as an “assistant” for the player providing hints and guidance. Finally, the story that guides the interactive chatbots is configurable through the Training Scenario Configurator (TSC) module so that tailored, inexhaustive scenarios are accommodated in the storytelling through a no-code environment for non-experts.
LAW-GAME has been a great opportunity for helvia.ai to expand its chatbots’ capabilities and automation offering. Moving beyond the usual implementations for chat agents that target at discovering the ‘correct’ information and serving it to the user, the LAW-GAME chat agents allow for interactive storytelling, providing alternative responses and reactions based on player’s input and thus, delivering a unique and realistic experience every time the game is played. In the implementation, we used an innovative combination of two approaches; a pre-conditions mechanism with alternative responses based on variables, and a response extraction mechanism through prompt engineering. Furthermore, we built a scenario editor to facilitate this combination of approaches, offering in this way a dynamic solution to scenario authors who may now author part of the storytelling through simple prompts (e.g., the truthful responses or known facts) and part through pre-conditions and variables (e.g., the responses that relate to the key evidences).
Moreover, LAW-GAME offered the ideal environment to extensively test and optimize voice and VR features for the chat agents. We have implemented Speech-To-Text and Text-To-Speech services and we had the chance to work with 3D environments in Unity as an additional channel for chatbot implementations including working with VR objects and scenes. Additionally, we introduced ‘Actions’ in the scenario editing, which are essentially instructions that pass through to the VR environment in Unity for triggering an in-game event, e.g., instructing the suspect avatar to perform a certain movement based on the response’s emotion.
For the last three years, Helvia.ai team had the pleasure to be part of a consortium of valuable partners that are dedicated at delivering innovative, high-quality solutions.