This paper surveys user opinions on issues related to autonomy in intelligent environments. The main contributions can be categorized into three parts: a great comparison with three previous studies, a presentation of quantitative and qualitative results from the questionnaire, and details explaining why users see great value in adjusting the autonomy of intelligent systems for real-life applications. The introduction makes it clear that the paper is positioned in between two extreme approaches to system design: end-user driven and autonomous-agent driven. Thus, this work assumes that users should be in charge of adjusting the level of control of the intelligent environments they manage, allowing the system to share tasks with humans and achieve a final outcome together. Three similar studies are referenced in order to better position the contributions of the paper within the context of managing control over intelligent environments. The three studies include an interview study that aimed to understand people’s concerns and requirements for smart homes, an exploratory study on the link between people and home technology, and a three-year project that investigated the use of proactive technology in the home.

The main section of the paper presents the survey. It discusses four management styles for intelligent environments: one fully autonomous-agent-driven style, two semi-autonomous (one high autonomy and one low autonomy) styles, and one fully end-user-driven style. Participants were asked how much autonomy they would allow in smart homes under the various management styles. The results represent data from 159 respondents over a three-month period. The individuals were asked, for example, to rate the usefulness of each management style on a five-point scale. From these responses, the authors present important findings:

• More than 80 percent of respondents found the fully end-user-driven style to be the most useful;
• The vast majority of users found the ability to change the management style either useful or very useful; and
• People have different autonomy needs for the different subparts of the system.

Altogether, the results support the main idea presented in the beginning, that is, the need for adjustable autonomy in intelligent environments. In conclusion, the results of this survey have strong implications for the design of agents in intelligent environments.