Abstract:

The Narrative Primer (NP) is an innovative, screenless educational device designed to revolutionize learning for children, especially those who are visually impaired or whose parents prefer audio-based educational media. Derived from post-smartphone Personal Primer artefact series, NP leverages advanced embedded-AI technologies to facilitate interaction through speech without necessity to connect to the cloud. The NP targets all children with a preference for auditory learning, focusing initially on those aged 6-11. Two primary groups are prioritized during the growth phase: blind or visually impaired pupils and children whose parents or communities are critical of screen-based media.

Introduction:
The Narrative Primer sits unassumingly on a pupil's desk, shelf, or by a window, where it can recharge using solar energy. At its core, the NP integrates modern iterations of Nvidia Jetson Nano systems, providing the computational power necessary for its operations. The primary mode of interaction with the NP is through speech, offering an engaging, screen-free experience.

At designated times, the NP's artificial Narrator activates and inquires if there is anyone present. If a child responds, the NP employs a voice identification algorithm to recognize the student. For new users, the device creates a "base profile" by asking introductory questions about preferences such as favorite colors, animals, and toys. Concurrently, it assesses the child's skills through non-intrusive methods.

Once the base profile is established, the Narrator begins storytelling, tailoring narratives based on the child's profile and competence level. This personalized approach is made possible by a 7-billion parameter large-language model installed on the NP device itself. The stories are crafted using curated templates from cultural treasures like Aesop’s fables, the Panchatantra, and Grimm's fairy tales, aligned with specific learning goals.

Curricular Goals and Learning Graph:
The NP’s educational framework, the Learning Graph (LG), is a structure akin to a probabilistic finite-state automaton, comprising learner states (vertices) and transition tasks (edges). These tasks—categorized as "listen," "imitate," "narrate," "test," and "feedback"—guide the learner towards mastering the curriculum. Each learner state has an associated knowledge unit, and progress is tracked by the child's level of mastery, which determines the complexity of subsequent stories and tasks.

In a feedback-loop mediated by story-telling, NP acquires information about the student while student acquires knowledge and competences which bring him/her closer to the curricular goal.

Human-Machine Narrative Learning:
NP emphasizes learning by narration, encouraging children to listen, repeat, and create stories. Each story is a blend of the child's profile, knowledge state, and a relevant story template. During storytelling, the NP prompts the child to repeat phrases, answer questions, or solve puzzles, providing immediate feedback and adjusting the narrative accordingly.

In sum, the NP seeks to revive the traditional educational practice of storytelling, enhanced by modern speech-to-text (STT), text-to-speech (TTS) and large-language model (LLM) technologies. This method fosters a deep, engaging learning experience away from the adverse effects of excessive screen time.

At ICERI, we will present a NP prototype addressing the Curricular Goals in fields of Basic Literacy, Arithmetics & Music.

Keywords:

Personal primer, narrative education, artificial intelligence in education, human-machine peer learning, speech-based technologies.