BIBLIOSMIA: Hyper-Personalized Consistent Stories for Enhanced Social Emotional Learning

Akib Shahriyar, Radwa Hamed, E. Margaret Perkoff, Mostafa Aboelnaga, Alya Azab
Proceedings of the Innovation and Responsibility in AI-Supported Education Workshop, PMLR 273:105-115, 2025.

Abstract

Consistent visual storytelling plays a central role in how humans teach their children to understand their emotions, relationships with others and the world around them. It enhances children’s cognitive and social-emotional development by providing engaging, believable narratives that help them navigate emotional complexities in a safe, imaginative context. This prevalence of visual storytelling in our lives has made it a prime application for technological advancements in artificial intelligence (AI). With AI integrations, digital stories can be created across an infinite set of topics and readily adapted to personalized contexts. As image generation algorithms advance, digital storytelling can be enhanced even further to incorporate visual elements that are unique to the author or the desired reader population. However, the burgeoning field of multi-modal story generation currently suffers from the problem of consistency - a critical element for preserving the plot lines of a story and the ability of children to relate to the characters therein. To mitigate this, we propose Bibliosmia, a novel framework for designing consistent digital stories. The framework encompasses three main components: a story generation module, an alignment module and an image generation and validation module that collectively preserve key narrative and visual story elements to allow expert and new authors alike to craft deeply personal developmental stories for children. We evaluated the effectiveness of the framework in the context of an online automatic story generation application. Our experimental results demonstrate Bibliosmia’s superior performance in prompt similarity (0.307 CLIP Score) and near-top-tier identity consistency (0.860 CLIP Score), surpassing other approaches in scalability and user satisfaction. These findings highlight Bibliosmia’s effectiveness in delivering high-quality, personalized storytelling experiences, setting a new standard in multi-modal digital storytelling.

Cite this Paper

BibTeX
@InProceedings{pmlr-v273-shahriyar25a, title = {BIBLIOSMIA: Hyper-Personalized Consistent Stories for Enhanced Social Emotional Learning}, author = {Shahriyar, Akib and Hamed, Radwa and Perkoff, E. Margaret and Aboelnaga, Mostafa and Azab, Alya}, booktitle = {Proceedings of the Innovation and Responsibility in AI-Supported Education Workshop}, pages = {105--115}, year = {2025}, editor = {Wang, Zichao and Woodhead, Simon and Ananda, Muktha and Mallick, Debshila Basu and Sharpnack, James and Burstein, Jill}, volume = {273}, series = {Proceedings of Machine Learning Research}, month = {03 Mar}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v273/main/assets/shahriyar25a/shahriyar25a.pdf}, url = {https://proceedings.mlr.press/v273/shahriyar25a.html}, abstract = {Consistent visual storytelling plays a central role in how humans teach their children to understand their emotions, relationships with others and the world around them. It enhances children’s cognitive and social-emotional development by providing engaging, believable narratives that help them navigate emotional complexities in a safe, imaginative context. This prevalence of visual storytelling in our lives has made it a prime application for technological advancements in artificial intelligence (AI). With AI integrations, digital stories can be created across an infinite set of topics and readily adapted to personalized contexts. As image generation algorithms advance, digital storytelling can be enhanced even further to incorporate visual elements that are unique to the author or the desired reader population. However, the burgeoning field of multi-modal story generation currently suffers from the problem of consistency - a critical element for preserving the plot lines of a story and the ability of children to relate to the characters therein. To mitigate this, we propose Bibliosmia, a novel framework for designing consistent digital stories. The framework encompasses three main components: a story generation module, an alignment module and an image generation and validation module that collectively preserve key narrative and visual story elements to allow expert and new authors alike to craft deeply personal developmental stories for children. We evaluated the effectiveness of the framework in the context of an online automatic story generation application. Our experimental results demonstrate Bibliosmia’s superior performance in prompt similarity (0.307 CLIP Score) and near-top-tier identity consistency (0.860 CLIP Score), surpassing other approaches in scalability and user satisfaction. These findings highlight Bibliosmia’s effectiveness in delivering high-quality, personalized storytelling experiences, setting a new standard in multi-modal digital storytelling.} }
Endnote
%0 Conference Paper %T BIBLIOSMIA: Hyper-Personalized Consistent Stories for Enhanced Social Emotional Learning %A Akib Shahriyar %A Radwa Hamed %A E. Margaret Perkoff %A Mostafa Aboelnaga %A Alya Azab %B Proceedings of the Innovation and Responsibility in AI-Supported Education Workshop %C Proceedings of Machine Learning Research %D 2025 %E Zichao Wang %E Simon Woodhead %E Muktha Ananda %E Debshila Basu Mallick %E James Sharpnack %E Jill Burstein %F pmlr-v273-shahriyar25a %I PMLR %P 105--115 %U https://proceedings.mlr.press/v273/shahriyar25a.html %V 273 %X Consistent visual storytelling plays a central role in how humans teach their children to understand their emotions, relationships with others and the world around them. It enhances children’s cognitive and social-emotional development by providing engaging, believable narratives that help them navigate emotional complexities in a safe, imaginative context. This prevalence of visual storytelling in our lives has made it a prime application for technological advancements in artificial intelligence (AI). With AI integrations, digital stories can be created across an infinite set of topics and readily adapted to personalized contexts. As image generation algorithms advance, digital storytelling can be enhanced even further to incorporate visual elements that are unique to the author or the desired reader population. However, the burgeoning field of multi-modal story generation currently suffers from the problem of consistency - a critical element for preserving the plot lines of a story and the ability of children to relate to the characters therein. To mitigate this, we propose Bibliosmia, a novel framework for designing consistent digital stories. The framework encompasses three main components: a story generation module, an alignment module and an image generation and validation module that collectively preserve key narrative and visual story elements to allow expert and new authors alike to craft deeply personal developmental stories for children. We evaluated the effectiveness of the framework in the context of an online automatic story generation application. Our experimental results demonstrate Bibliosmia’s superior performance in prompt similarity (0.307 CLIP Score) and near-top-tier identity consistency (0.860 CLIP Score), surpassing other approaches in scalability and user satisfaction. These findings highlight Bibliosmia’s effectiveness in delivering high-quality, personalized storytelling experiences, setting a new standard in multi-modal digital storytelling.
APA
Shahriyar, A., Hamed, R., Perkoff, E.M., Aboelnaga, M. & Azab, A.. (2025). BIBLIOSMIA: Hyper-Personalized Consistent Stories for Enhanced Social Emotional Learning. Proceedings of the Innovation and Responsibility in AI-Supported Education Workshop, in Proceedings of Machine Learning Research 273:105-115 Available from https://proceedings.mlr.press/v273/shahriyar25a.html.

Related Material