Jahad Daneshgahi Prioritizes Development of Indigenous AI Models

The head of the AI Research and Advancement Center, Afshin Sandoughdar, at Jahad Daneshgahi (Academic Center for Education, Culture, and Research - a major public research and academic organization in Iran.) has announced the launch of a project to develop deep Persian-language models, emphasizing that the initiative aims to better capture the cultural and literary nuances of Iranian society.
 

In an interview with ISNA (News Agency in Iran), Afshin Sandoughdar highlighted the sensitivity of applying artificial intelligence in cultural and social domains, stating that defining the boundaries of its use is among the center’s key strategic missions. He noted that relying solely on quantitative models and pre-defined algorithms is insufficient for making definitive judgments about complex cultural phenomena.

Referring to the center’s four guiding principles in applying AI to cultural and social issues, he explained that the first principle is localization and contextualization. According to him, no algorithm or model will be used in research without being adapted to Iran’s cultural, linguistic, and social context, as many existing tools are built on Western data and assumptions that may lead to misleading conclusions if applied directly.

Sandoughdar identified interdisciplinarity as the second principle, stressing that no cultural or social project will be approved or implemented without the active involvement of experts in the humanities and social sciences. Specialists such as sociologists, anthropologists, social psychologists, and communication experts play a crucial role in preventing oversimplification and ensuring that research methods align with the complexity of cultural phenomena.

He added that the third principle focuses on social participation and field validation. Research findings, he said, must be shared with target groups and stakeholders so that real-world feedback can be incorporated into the analysis and refinement of models. While algorithms may reveal correlations, interpreting them within lived social contexts requires human expertise.

Emphasizing the fourth principle, Sandoughdar stated that transparency and explainability are essential in cultural research. Researchers must clearly indicate which parts of an analysis are conducted by AI, what algorithms and assumptions are used, and where human judgment plays a role.
 

He also outlined plans for collaboration with academic institutions, expressing the need for closer cooperation with the Research Institute for Humanities and Social Studies of Jahad Daneshgahi in the coming year. One key initiative involves designing a system to intelligently monitor cultural trends in cyberspace. Using Persian language processing and machine learning, the system will aim to identify shifts in cultural preferences, lifestyles, and media consumption patterns among Iranians. According to Sandoughdar, cultural studies experts will work alongside computer engineers throughout the project to oversee algorithm design and interpret findings.
 

He further announced a joint project in sociology for “Modeling Social Dynamics and Social Capital.” The initiative will combine survey data with information derived from online social interactions to develop a framework for measuring and predicting changes in social capital across different regions of the country.
He emphasized that the continuous involvement of sociologists will ensure accurate operationalization of social capital indicators and enable analysis that goes beyond purely quantitative approaches toward a deeper understanding of qualitative social dimensions.
 

Mr Sandoughdar also pointed to expanding scientific collaborations in areas such as language, cultural economics, and international engagement. He stressed that the center’s approach to AI is based on localization, collective expertise, and interdisciplinary capacity. In this context, partnerships with Persian language and literature departments at major universities, including the University of Tehran and Tarbiat Modares University, are on the agenda.
 

Highlighting the unique challenges of Persian language processing due to its specific morphological and syntactic features, Mr Sandoughdar announced that the “Deep Language Models for Persian” project will begin soon, aiming to create advanced models capable of more accurately understanding the subtleties of the language.

News Report: https://www.isna.ir/news/1405021206874/توسعه-مدل-های-بومی-هوش-مصنوعی-برای-زبان-فارسی-در-دستور-کار-جهاد