Response time in man-computer conversational transactions

Report Summary

In 1968 IBM's Robert Miller conducted exhaustive research into how humans interact with computers. Specifically he was interested in the different range of responses that occur while waiting for a computer to respond to a human request.

His study looked at many different types of human computer interaction and he established that we work either conversationally or in a transactional manner.

He identified that if humans have to wait too long for a response then any conversational-based interaction became interrupted and was therefore considered transactional in nature.

In transactional behavior, humans disconnect from the computer to conduct other activities and tasks and may even lose complete interest altogether.

Consequently, to remain conversational from a human perspective, the shorter the interaction time the better. Observation also found that if the response time became too short, it may go unnoticed and it could induce further reaction time while recognition of a response had occurred.

He concluded that at about 1s, responsiveness was considered seamless and enables human interaction to occur un-interrupted. The optimum response time was considered 3s as after this time he observed that a move to transactional interaction increased with almost everyone switching away from conversation at 10s.

Currently best practice is to initiate the initial render to the device viewport (screen), within 0.75s and complete the full page download within 3 seconds.

When this research was conducted technology was fundamentally simpler. However, the principles and observations of Robert Miller's research is still of interest as it underpins the foundation for some of our current approach and thinking of how a high quality user experience should be delivered.