U-II 98.18-23 APRIL 1998 PAPERS Persuasive Computers: Perspectives and Research Directions BJ Fogg Stanford University Box 8333, Stanford, CA 94309 USA bjfogg@stanford.edu, bjfogg@acm.org www.captology.org ABSTRACT The study of computers as persuasive technologies (called ���captology���) was introduced at CHI 97 as a new area of inquiry. This paper proposes definitions, perspectives, and research directions for further investigation of this field A persuasive computer is an interactive technology that attempts to change attitudes or behaviors in some way. Perspective 1 describes how computers can inherit three types of intentionality: endogenous, exogenous, and autogenous. Perspective 2 presents the ���Functional Triad,��� which illustrates that computers can function as persuasive tools, media, and social actors. Perspective 3 presents a ���levels of analysis��� approach for captology, which include varying levels from individual to societal- Perspective 4 suggests a simple method for exploring the design space for persuasive computers. Perspective 5 highlights some ethical issues inherent in persuasive computing. The paper concludes by proposing seven directions for further research and design. Keywords persuasion, captology, media, computers as social actors, ethics, design methods, computers as persuasive technologies INTRODUCTION At CHI 97, a special interest group meeting gathered a number of participants who were interested in exploring the domain of computers and persuasion [a- We agreed to call this area ���captology��� (built from an acronym for Computers As Persuasive Technologies), with the graphic in Figure 1 serving as a reference point for this domain. The discussion on captology at the CHI 97 SIG proved fruitful and enlightening, with participants concurring that captology was an intriguing area for further research and design. The group also agreed that this domain had not yet been adequately defined or addressed by researchers and practitioners of human- computer interaction. We found that our discussion suffered at times because we lacked key definitions and frameworks for understanding captology. The purpose of this paper, therefore, is to contribute to the CHI community���s understanding of persuasive computing by proposing definitions, perspectives, and research directions for the field of captology. l���cnnission to make digimlkurl copies ofzdl or pti of this msterial for pxsonal or clssroom use is gmnted witbout fee provided that the copies are not made or diibuted for profit or commcrcird advantqe, the copy- rightnotic+tbetitleoftbepubliwtionand its&teappear,andnoticek given that copyright is by permission oftbeACh% Inc. To copy otbentise, lo rcpubliih, lo post,on servers or to redistribute to Iii requires -qecific p~mtission andlorfez CHI 9S Los Angeles CA USA copyright 199s o-g9791-975-o19W 4.s.00 Sun Microsystems 901 San Antonio Road, MPK17-105 Palo Alto, CA 94303 USA b.j.fogg@sun.com Figure 1: Captology describes the shaded area where computing technology and persuasion overlap. This paper first presents five perspectives on computers and persuasion. They include the following: Perspective 1: Definition of persuasive computers Perspective 2: A functional view of persuasive computers Perspective 3: Levels of analysis for captology Perspective 4: The design space for persuasive technologies Perspective 5: Ethics of computers that persuade Each of these five perspectives provides a different way to view persuasive computers, while also describing examples of relevant technologies. After setting forth the five perspectives on persuasive technologies, this paper concludes by outlining seven profitable directions for further research in the area of persuasive computers. To be clear, this paper makes a contribution by articulating a range of approaches to captology. It is the role of later work to expand and revise the ideas proposed in this paper. PERSPECTIVE 1: DEFINITION OF PERSUASIVE COMPUTERS What is a persuasive computer? Simply put, a persuasive computer is an interactive technology that changes a person���s attitudes or behaviors. This definition works well in many cases, but a more thorough definition gives a better understanding of persuasive computing. The psychology literature suggests many definitions for the word ���persuasion��� [e.g., 24, 361. After reviewing the work of persuasion scholars, I���ve synthesized the various definitions to define ���persuasion��� as ���an attempt to shape, reinforce, or change behaviors, feelings, or thoughts about an issue, object, or action.��� 225

- - .A. -^ _ - --.- ,- . . . .ii ^r-.-- -...���---x^-.. _-,. ���.&:..-,.. -,_ I_-, PAPERS CHI 98 . 18-23 APRIL 1998 Persuasion and intentionality One key point in this definition is that true persuasion implies an intent to change attitudes or behaviors in other words, persuasion requires intentionality. Therefore, not all behavior or attitude change is the result of persuasion. For example, a rain storm may cause people to buy umbrellas, but the storm is not a persuasive event because it has no intentional@ associated with it. (However, if an umbrella manufacturer could somehow cause rain, then the rain storm might qualify as a persuasive tactic.) Because machines do not have intentions [S], a computer qualifies as a persuasive technology only when those who create, distribute, or adopt the technology do so with an intent to affect human attitudes or behaviors. To be clear, the persuasive nature of a computer does not reside with the object itself instead, a computer being classified as ���persuasive��� depends on the context of creation, distribution, and adoption. I propose that if an intent to change attitudes or behaviors is a factor in the creation, distribution, or adoption of a technology, then that technology inherits a type of intent from human actors. Three types of intent- endogenous, exogenous, and aufogenous For the purposes of captology, I propose three kinds of inherited persuasive intent: endogenous, exogenous, and autogenous. According to my definitions, a computer technology inherits endogenous intent when a designer or producer creates a technology with intent to persuade users in some way. A computer technology inherits exogenous intent when one person provides another person with a computer technology in an attempt to change that person���s attitudes or behaviors. A computer technology inherits autogenous intent when a person chooses to use or adopt a technology in order to change his or her own attitudes or behaviors. Table 1 makes this idea clearer. I Type of intent external factors��� autogenous ���self-produced Where intent comes Example from Those who create or Health-Hero video games are produce the designed to persuade interactive children to develop good technology health habits [IT& Those who give A mother may give her son a access to or Palm Pilot PDA in hopes that distribute the he will become more interactive organized. technology to others The person adopting A person may by and use a or using the calodecounting computer interactive device to help change his or technology her own eating behavior. Table 1: Three types of intent wfth examples Although the above categories aim to distinguish among types of persuasive technologies, I recognize that these categories are not always precise, and they are not always mutually exclusive. Indeed, making inferences about intentions is tricky business-we may infer intent where there is none, or we may fail to perceive intent when intent does indeed exist. Furthermore, it is quite possible that a given interactive technology may fall into more than one category. Despite the potential ambiguities, I find these three categories helpful in better understanding the range and roles of persuasive computing technologies. PERSPECTIVE 2: A FUNCTIONAL VIEW OF PERSUASIVE COMPUTERS While Perspective 1 provides an ���intentional��� framework for persuasive computers, Perspective 2 presents what I call a ���functional��� view. To explain Perspective 2 clearly, I first describe the basics of the ���Functional Triad.��� I then show how this framework can provide key insights into the study of persuasive computers. The Functional Triad I propose that today���s computers function in three basic ways: as tooIs, as media, and as social actors. Researchers and designers have often discussed variants of these functions [e.g., 18,22,33], usually as metaphors for computer use. However, I suggest that these three categories are more than metaphors they are basic ways that people view or respond to computing technologies. I refer to these three areas as ���functions.��� As a tool, the computer (or the computer application or system) provides humans with new ability or power, allowing people to do things they could not do before, or to do things more easily [28,29]. Computers also function as media [13, 30, 321, a role that has become more apparent and important in recent years [30, 341. As a medium, a computer can convey either symbolic content (e.g., text, data graphs, icons) or sensory content (e.g., real-time video, simulations, virtual worlds). Computers can also function as social actors [12, 16, 25, 321. Users seem to respond to computers as social actors [25] when computer technologies adopt animate characteristics (physical features, emotions, voice communication), play animate roles (coach, pet, assistant, opponent), or follow social rules or dynamics (greetings, apologies, turn taking). Mapping the funcfions One way to view these three functions simultaneously is to map the three categories into two dimensions. Of course, in all but the most extreme cases, these functions mix and blur in any one given interactive technology. However, consciously mapping interactive technologies into a triangular space I call the ���Functional Triad��� generates insight into the roles and relationships of different computer technologies. Figure 2 represents the Functional Triad with some prototypical examples. Persuasion and the functional view of computers Although the functional triad is a useful framework for understanding computer technologies in general, adopting this functional view of computers yields specific insights for analyzing persuasive interactive technologies. 226