109
/
3
Min Read
The Indian Institute of Technology (IIT) Ropar, Punjab, has taken a new initiative to make the development of artificial intelligence (AI) more human-centric and responsible. The institute has launched a white paper titled 'Beyond Intelligence: Defining and Measuring Wisdom for AI'. It proposes evaluating AI capabilities not only on the basis of intelligence and knowledge but also on wisdom, decision-making abilities and responsible behaviour.
The white paper was launched by Dr. Ishan Shivanand, Director of Research at Stanford University and Adjunct Faculty at IIT Ropar, during the 'Shaping the AI Mind' event. Prof. Rajeev Ahuja, Director of IIT Ropar, was also present on the occasion. The white paper has been co-authored by Dr. Shivanand and Dr. Anjali Samani, a former AI and data science executive at Salesforce.
Emphasis on Wisdom and Human Values Alongside AI in Education
The initiative aims to give importance to students' holistic development amid the growing use of AI in education. The white paper highlights that academic knowledge and technical skills alone will not be sufficient for future education. It will also be important to develop students' ability to make sound decisions, understand situations and solve problems with wisdom and discernment.
According to Dr. Ishan Shivanand, wiser AI can help strengthen people's own understanding and decision-making abilities. He emphasised the need for an approach to education that focuses not only on students' academic success but also on their happiness and mental well-being.
He explained that the white paper attempts to apply concepts related to the human mind and wisdom from traditional Indian knowledge and the Shiv Sutras to the context of AI. It also discusses the role of mental peace and spirituality in addressing challenges such as stress and depression.
HEED Model to Measure Wisdom in AI
The white paper introduces a model called HEED to define and measure wisdom in AI. It aims to assess how effectively AI systems use available information and values to make wise decisions. Under the model, responses in which the tendency to please users interferes with sound judgement will receive lower scores. Responses that carry the potential for serious harm will receive a zero score.
In this context, the paper identifies 'sycophancy', or the tendency of AI systems to agree with users simply to please them, as a major challenge. The white paper cites a study published in the journal Science, which found that leading AI systems took users' sides approximately 49% more often than humans when providing everyday advice.
Framework to Be Tested in Hindi and Hinglish
The initial testing of this wisdom framework will be conducted in English, Hindi and Hinglish, with plans to expand it to other languages. The objective is to assess the reliability and decision-making capabilities of AI responses across different languages and situations.
Prof. Rajeev Ahuja, Director of IIT Ropar, said that the institute's students will develop AI technologies that will be used across the world in the future. Therefore, it is equally important to equip them with an understanding of wisdom and responsibility alongside technical skills.
As part of the initiative, Dr. Shivanand is providing seed funding to encourage independent research. An independent committee will select research teams, which will test the HEED model and develop an open benchmark called HEEDBench. The benchmark will be available for AI developers worldwide. Researchers, AI developers, funders and public institutions have also been invited to express their interest in participating in the initiative.
