Multimodal Large Language Models as Synthetic Participants in Video-Based Studies: An Evaluation
arXiv:2606.07541v1 Announce Type: cross Abstract: Multimodal large language models (MLLMs) have shown strong performance on objective tasks such as video understanding and reasoning. However, it remains unclear whether they can approximate subjective human responses, which depend not only on content comprehension but also on individuals' social contexts. To address this gap, we evaluate MLLMs as synthetic participants in an emerging task: assessing perceived sensory engagement with short videos. Grounded in the Perceived Message Sensation Value (PMSV) framework, we compare ratings from recruit
The rapid advancement and sophistication of Multimodal Large Language Models (MLLMs) are enabling them to not only understand objective information but also to simulate subjective human responses, pushing the boundaries of AI capabilities.
This development indicates MLLMs can move beyond simple data processing to emulate nuanced human perception and social context, impacting fields reliant on human feedback and potentially accelerating the development of truly autonomous AI agents.
MLLMs can now be considered as viable synthetic participants for certain types of qualitative research, offering a faster and more scalable alternative to human recruitment for studies assessing subjective experiences.
- · AI model developers
- · Market research agencies
- · Digital content creators
- · Social scientists
- · Traditional survey companies
- · Human participant recruitment services
- · Low-skilled data labelers
AI models will increasingly replace human participants in subjective research and evaluation tasks.
This will accelerate the iteration and development cycles for products and services that rely on user feedback.
The definition of 'human-like' intelligence will become increasingly blurred as AI systems demonstrate capabilities previously thought to be uniquely human.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI