ELEMENTAL PATH COGNITOYS: COMBINING RULE-BASED DIALOGUE MANAGEMENT, MODEL-TRACING AND EXAMPLE-TRACING INSTRUCTIONS TO DYNAMICALLY PERSONALIZE DIALOGUE-BASED LEARNING WITH TOPICS OF INTEREST
Elemental Path Inc. (UNITED STATES)
In overview, a conversational smart toy that holds intelligent conversations with young children is in itself a novelty. Many speech-enabled technologies such as Siri of Apple, Echo of Amazon and Cortana of Microsoft, are naturally language query engines that have little to no regard for continuous discourse or instructional scaffolding. Suffice to say, these products do not facilitate knowledge acquisition. Conversational toys of equivalent NLP capacity are practically nonexistent. As far as intelligent tutoring systems are concerned, young children often go off-task while working with a computer-based learning interface without gaming or role-play components. Thus, there is a compelling reason to create a connected smart toy to bridge natural language dialogue management with model-tracing and example-tracing instructional technologies to deliver an interesting and engaging learning experience for early childhood.
Elemental Path’s Cognitoys are a platform built specifically to address this challenge of creating conversational agents in physical toys that blend learning into highly personalized conversations.
For example, a child may engage by asking Cognitoy to tell the story of Snow White. After story-telling, the toy may ask the child questions pertaining to details of the story (e.g. how many dwarfs did Snow White meet?). A child may stray away from the conversation and begin talking about cats. Subsequently, the toy handles the new conversation by asking the child how many legs a cat has. The child answers with “four”. The toy then determines the appropriate level of math based on age an decides engage the child in a multiplication-by-counting by asking how many legs do five cats have in total. If the child responds incorrectly, the toy then guides the child through a scaffolded process of performing counting-by-four as a way to multiply. If the child succeeds in precisely demonstrating the skill, the toy will increase the difficulty of instructions by locating a relevant, more challenging goal. Additionally, if the child exhibits overwhelming off-task behavior, the toy is equipped to reengage by either resuming the conversation about Snow White, or initiating a new learning exercise such as finding words that rhyme with “cat” or “leg”. The toy will seek to reinitiate math instructions later.
The approach that Cognitoys takes is a hybrid conversational system that employs both non-finite-state, rule-based dialogue management (like RavenClaw/Olympus), and model-tracing (like ACT-R) and ex-ample-tracing instructions (like CTAT). Rule-based dialogue manage-ment allows flexible continuation of casual, themed conversations; whereas instructional tracing engine uses content of interest to the student collected through casual conversations to dynamically personalize assessments and instructional scaffolding without compromising the integrity of instructional designs. Such approach enables Cognitoys to optimize a child’s learning experience by creating and managing educational conversations that continually stimulate student interest in real-time.
(A intro video based on a pilot test can be found here: https://www.kickstarter.com/projects/522717158/cognitoys-internet-connected-smart-toys-that-learn)