Users generally trust computer interfaces to accurately re- T 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.

plored features that enable this type of deception. While research into human-to-human deception has advanced our understanding of deception in general, it has largely focused on communication behavior that pre-exists, and per- sists through, computermediated-communication.

Most of the systematic research in the academic HCI com- munity focuses on malevolent deception, or deception in- tended to benefit the system owner at the expense of the user [14]. Such research frames deception negatively, and focuses on detection and eradication of malicious or evil interfaces (e.g., dark-patterns [17]). Such patterns include ethically dubious techniques of using purposefully confus- ing language to encourage the addition of items to a shop- ping cart or hiding unsubscribe functionality. Many forms of malevolent deception, such as phishing and other fraud, is in clear violation of criminal law.

Deceptive practices that are considered harmful by legal organizations are, for generally good reasons, considered harmful by designers as well. A possible exception to the negative frame for malevolent deception is in contexts where an obvious adversary exists or for military or securi- ty purposes. For example, to detect network attacks or illicit use, security researchers may deploy “honeypots” to imitate vulnerable

plored features that enable this type of deception. While research into human-to-human deception hcceptable even when it might affect legitimate clients who try to connect even though no services are available.

It can be difficult to draw the line between malevolent and benevolent deception. We frame the distinction from the end-user’s perspective: if the end-user would prefer an ex- perience based on the deceptive interface over the experi- ence based on the “honest” one, we consider the deception benevolent. Note that this includes situations in which both the system designer and the end-user benefit from the de- ception, something economists sometimes term a Pareto white lie [19]. Arguably, all benevolent deceptions that im- prove user experience benefit the interface creator as well; otherwise, we might use the term altruistic deception [19].

The limited HCI research that has been conducted on be- nevolent deception has focused on the use of magic [54], cinema and theater [32] as instructional metaphors for HCI design. By using deception for the purpose of entertain- ment, playfulness, and delight, it becomes acceptable. This work has connected these art forms, in which illusion, im- mersion, and the drive to shape reality dominate, to situa- tions in which there is willing suspension of disbelief on the part of the user, or at least willing pursuit of acceptable mental models. Similar lessons have been drawn from ar- chitecture. For example, theme parks and casinos [20, 34, 37] are designed specifically to utilize illusions that manip- ulate users’ perception of reality, entertainment, and partic- ipation in the experience. In this paper, we will describe a number of deceits in HCI systems that have parallel designs to magic, theater, and architecture. However, not all benevolent deceit can be framed in terms of creating delight for as advanced our understanding of deception in general, it has largely fo- cused on communication behavior that pre-exists, and per- sists through, computer-mediatedcommunication. Most of the systematic