Arash Eshghi

Heriot-Watt University, Heriot-Watt University Malaysia

Papers

Total Citations

H-Index

About

Arash Eshghi is a researcher at the forefront of human-robot interaction and explainable AI, with a primary focus on how machines can learn and communicate through natural language. His work centers on developing dialogue systems that are not only interactive but also capable of grounding new word meanings in visual contexts, a critical step toward more transparent and adaptable AI. Eshghi’s most impactful contribution is his exploration of automatic evaluation metrics for natural language explanations, a paper with 36 citations that bridges the gap between transparency in robotics and the established field of NL generation. This work addresses a pressing need: as AI systems become more autonomous, the ability to generate and assess clear, human-understandable explanations is paramount. He also developed VOILA, an optimised multimodal dialogue agent that learns visually-grounded word meanings interactively from users, demonstrating a practical application of his research. By training on real human-human dialogues, VOILA achieves a level of natural interaction that pushes the boundaries of how robots can learn from and communicate with people. Eshghi’s research is essential for anyone interested in building AI that is both intelligent and accountable.