Users generally trust computer interfaces to accurately re- flect system state. Reflecting that state dishonestly— through deception—is viewed negatively by users, rejected by designers, and largely ignored in HCI research. Many believe outright deception should not exist in good design. For example, many design guidelines assert: “Do not lie to your users” (e.g., [40, 45]) Misleading interfaces are usually attributed to bugs or poor design. However, in reality, de- ceit often occurs both in practice and in research. We con- tend that deception often helps rather than harms the user, a form we term benevolent deception. However, the over- loading of “deception” as entirely negative coupled with the lack of research on the topic, makes the application of de- ception as a design pattern problematic and ad hoc.

Benevolent deception is ubiquitous in real-world system designs, although it is rarely described in such terms. One example of benevolent deception can be seen in a robotic physical therapy system to help people regain movement following a stroke [8]. Here, the robot therapist provides stroke patients with visual feedback on the amount of force they exert. Patients often have selfimposed limits, believ- ing, for example, that they can only exert a certain amount of force. The system helps patients overcome their percep- tive limits by underreporting the amount of force the patient actually exerts and encouraging additional force.

The line between malevolent and benevolent deception is fuzzy when the beneficiary of the deception is ambiguous. For example, take the case of deception in phone systems to mask disruptive failure modes: The connection of two indi- viduals over a phone line is managed by an enormous spe- cialized piece of hardware known as an Electronic Switch- ing System (ESS). The first such system, the 1ESS, was designed to provide reliable phone communication, but given the restrictions of early 1960s hardware, it sometimes had unavoidable, though rare, failures. Although the 1ESS knew when it failed, it was designed to connect the

As is the case with the 1ESS and placebo buttons, deception sometimes benefits the system designer, service provider, or business owner. However, this does not invalidate the fact that it might also help meet user needs. We believe that by not acknowledging that there is deception, and, more critically, that a line between beneficial and harmful deceptions might exist, research in the area is difficult to pursue—to the detriment of academics and practitioners alike.

A further example of benevolent deception are the “placebo buttons” that allow users to feel as though they have control over their environment when they actually do not. Cross- walk buttons, elevator buttons, and thermostats [33, 47] often provide no functionality beyond making their users feel as though they can affect their environment. Some of these buttons go far to provide the illusion of control; non- working thermostat buttons, for example, are sometimes designed to hiss when pressed [2]. In addition to providing the feeling of control, placebo buttons can signal the exist- ence of a feature to the user. Non-working crosswalk buttons, for example, clearly convey to a pedestrian that a crosswalk exists.

Whether intentional or not, implicit or explicit, acknowl- edged or not, benevolent deceit exists in HCI. Nonetheless, little is known about the motivation, mechanisms, detecta- bility, effectiveness, successes, failures, and ethics of this type of deception. Researchers have tiptoed around this taboo topic, concentrating instead on malevolent deception (e.g., malware or malicious software [14,17]) and unobjec- tionable forms of deception described using entertainment metaphors (e.g., magic or theater [32,54]). This limited view of deception does not capture its variety or ubiquity.

As we will see, one of the underlying reason for the ubiqui- ty of deception is that it can fill the many of the gaps and tensions that emerge with different design concerns (e.g., the good of the individual versus the good of the group), design goals (e.g., conflicting principles), or systems states (e.g., desired system performance versus actual system per- formance). In any situation where a poor fit exists between desire (e.g., the mental model or user expectations) and reality (e.g., the system itself) there is an opportunity to employ deception. This gap—which is extremely com- mon—both motivates and