Return to Beyond-Intelligent-Machines CVS log

Up to [Micropolis - Activity] / micropolis / src / notes

Annotation of micropolis/src/notes/Beyond-Intelligent-Machines, revision 1.1

1.1     ! root        1: Date: Tue, 24 Nov 92 22:25:59 -0500
        !             2: From: [email protected] (Ben Shneiderman)
        !             3: To: [email protected], [email protected], [email protected],
        !             4:         [email protected], [email protected]
        !             5: Subject: Re: more pie menus!
        !             6: 
        !             7: I couldn't resist sending you all this latest essay which is
        !             8: destined for IEEE Software...some readers expect it to generate
        !             9: some strong responses...Ben
        !            10: 
        !            11: 
        !            12: Beyond Intelligent Machines:
        !            13:     Designing Predictable and Controllable User Interfaces
        !            14:  
        !            15:  
        !            16:   Ben Shneiderman  November 24, 1992
        !            17:  
        !            18:        University of Maryland, College Park, MD 20742  
        !            19:  
        !            20:    Professor, Department of Computer Science,
        !            21:    Head, Human-Computer Interaction Laboratory at the 
        !            22:       Center for Automation Research &
        !            23:    Member, Institute for Systems Research
        !            24:  
        !            25:  
        !            26: Who's in control?
        !            27: 
        !            28: An important shift is occurring from the old vision of computers
        !            29: as 'intelligent' to a new vision based on predictable and controllable
        !            30: user interfaces that depend on direct manipulation of objects and actions.
        !            31: Appropriate metaphors and terminology are important since they shape
        !            32: the thoughts of researchers, designers, managers, congress-people, 
        !            33: journalists, etc.  Most of us have learned the importance of gender 
        !            34: neutral terminology and similarly I have been strongly opposed to 
        !            35: suggesting that computers are 'intelligent' or 'smart' for several 
        !            36: reasons:
        !            37:  
        !            38: 1) Limits to Imagination
        !            39:  
        !            40: I think we should have much greater ambition than to make a computer 
        !            41: behave like an intelligent butler or other human agent.  Computer 
        !            42: supported cooperative work (CSCW), hypertext/hypermedia, multi-media, 
        !            43: information visualization, and virtual realities are powerful 
        !            44: technologies that enable human users to accomplish tasks that no human 
        !            45: has ever done.  If we describe computers in human terms then we
        !            46: run the risk of limiting our ambition and creativity in the design
        !            47: of future computer capabilities.
        !            48:  
        !            49:  
        !            50: 2) Predictability and Control are Desirable
        !            51:  
        !            52: If machines are 'intelligent' or 'adaptive' then they may become less 
        !            53: predictable and controllable.  Our usability studies show that users 
        !            54: want feelings of mastery, 
        !            55: competence, and understanding that come from a predictable and 
        !            56: controllable interface.  Most users seek a sense of 
        !            57: accomplishment at the end of the day, not the sense that this 
        !            58: 'intelligent' machine magically did their job for them. 
        !            59:  
        !            60:  
        !            61: 3) Human Responsibility
        !            62:  
        !            63: I am concerned that if designers are successful in convincing the users 
        !            64: that computers are intelligent, then the users will have a reduced sense 
        !            65: of responsibility for failures.  The tendency to blame the machine is 
        !            66: already widespread and I think we will be on dangerous grounds if we 
        !            67: encourage this trend.
        !            68:  
        !            69:  
        !            70: 4)  Machines are not People  AND  People are not Machines
        !            71:  
        !            72: I have a basic philosophical objection to the suggestion that machines 
        !            73: are, or can ever be, intelligent.  I know that many of my colleagues are 
        !            74: quite happy to call machines intelligent and knowledgeable, but I prefer 
        !            75: to treat and think about machines in very different ways from the way I 
        !            76: treat and think about people.
        !            77:  
        !            78: 
        !            79: The lessons of history
        !            80:  
        !            81: While some productive work has been done under the banner of
        !            82: `intelligent', often those who use this term reveal how little they
        !            83: know about what users want or need.  The users's goal is not to
        !            84: interact with an 'intelligent' machine, but to create, communicate, 
        !            85: explore, plan, draw, compose, design, or learn.  Ample evidence 
        !            86: exists of the misguided directions brought by 'intelligent' machines:
        !            87:  
        !            88:   - natural language interaction seems clumsy and slow compared to 
        !            89: direct manipulation and information visualization methods that use 
        !            90: rapid, high-resolution, color displays with pointing devices.  Lotus HAL 
        !            91: is gone, AI INTELLECT hangs on but is not catching on.  There are some 
        !            92: interesting directions for tools that support human work through 
        !            93: natural language processing: aiding human translators, parsing 
        !            94: texts, and generating reports from structured databases.
        !            95:  
        !            96:   - speech I/O in talking cars and vending machines is gone.
        !            97: Voice recognition is fine for handicapped users plus special situations, 
        !            98: but doesn't seem to be viable in general office, home, or school 
        !            99: settings.  Our recent studies suggest that speech I/O has a greater
        !           100: interference with short term and working memory than hand-eye 
        !           101: coordination for mouse menu selection.  Voice store and forward,
        !           102: phone-based information retrieval, and voice annotation have great 
        !           103: potential but these are not the 'intelligent' applications.
        !           104:  
        !           105:   - adaptive interfaces are unstable and unpredictable, often leading 
        !           106: users to worry about what will change next.  I see only modest chances 
        !           107: for success in user modeling to recognize the level of expertise and
        !           108: revise the interface accordingly - can anyone point to successful
        !           109: studies or commercial products?  By contrast, user controlled 
        !           110: adaptation through control panels, cruise control for cars, and 
        !           111: remote controls for TV are success stories.  While algorithms to 
        !           112: deal with dynamic
        !           113: issues in network or disk space management are needed, the task domain 
        !           114: and user interface issues of the application program
        !           115: should generally be under direct user control.
        !           116:  
        !           117:   - Intelligent CAI (Computer Assisted Instruction) only prolonged the 
        !           118: time (compared to traditional CAI) until the users felt they were the 
        !           119: victims of the machine.  Newer variations such as Intelligent Tutoring 
        !           120: Systems are giving way to Interactive Learning Environments where 
        !           121: students are in control and actively creating or exploring.
        !           122:  
        !           123:   - intelligent talking robots with five-fingered hands and human facial 
        !           124: features (quaint fantasy that did well in Hollywood but not in Detroit 
        !           125: or elsewhere) are mostly gone in favor of flexible manufacturing systems 
        !           126: that enable supervisors to specify behavior with predictable results.
        !           127:  
        !           128:  
        !           129: It seems that some designers continue to ignore this historical pattern 
        !           130: and still dream of creating 'intelligent' or 'smart' machines.  It is an 
        !           131: ancient and primitive fantasy, and its seems most new technologies must 
        !           132: pass through this child-like animistic phase.  Lewis Mumford identified 
        !           133: this pattern (Technics and Civilization, 1934) when he wrote about the 
        !           134: Obstacle of Animism: 'the most ineffective kind of machine is the 
        !           135: realistic mechanical imitation of a man or another animal...for 
        !           136: thousands of years animism has stood in the way of...development.'   
        !           137:  
        !           138:  
        !           139: An alternate vision
        !           140: 
        !           141: My point in this essay is not merely to counter a popular design 
        !           142: philosophy, but to offer a new vision that is more in harmony with what 
        !           143: users want.  I believe that the future will be filled with powerful, but 
        !           144: predictable and controllable computers that genuinely serve human needs  
        !           145: (Designing the User Interface: Strategies for Effective Human-Computer
        !           146: Interaction, Second Edition, Addison-Wesley Publ. Co., Reading, MA, 1992).
        !           147: 
        !           148: In this vision of predictable and controllable (PC) computing, 
        !           149: the promising strategies are rapid,
        !           150: visual, animated, colorful, high resolution interfaces built on
        !           151: meaningful control panels, appropriate preference boxes,
        !           152: user-selectable toolbars, rapid menu selection, easy to create macros,
        !           153: and comprehensible shortcuts.  These enable me to specify rapidly, 
        !           154: accurately, and confidently how I want my email filtered, what documents 
        !           155: I want retrieved and in what order, and how my documents will be 
        !           156: formatted.
        !           157:  
        !           158:  
        !           159: Our Human-Computer Interaction Laboratory has applied these principles 
        !           160: to information visualization methods that give users X-ray vision to see 
        !           161: through their mountains of data.  Treemaps enable users to see (and 
        !           162: hear) 2-3000 nodes of hierarchically structured information by utilizing 
        !           163: every pixel on the display.  Each node is represented by a rectangle 
        !           164: whose location preserves the logical tree structure and whose area is 
        !           165: proportional to one of its attributes.  Color represents a second 
        !           166: attribute and sound a third (B. Johnson & D. Turo,  Improving the 
        !           167: Visualization of Hierarchies with Treemaps: Design Issues and 
        !           168: Experimentation, Proc. IEEE Visualization '92).  Treemaps have been 
        !           169: applied to Macintosh directory browsing (Figure 1), in which area could 
        !           170: be set to file size, color to application type, and sound to file age 
        !           171: (our TreeViz application is available from the University of Maryland's 
        !           172: Office of Technology Liaison, (301) 405-4210).  When users first try 
        !           173: TreeViz they usually discover duplicate or misplaced files, redundant 
        !           174: and chaotic directories, and many useless files or applications.  Other 
        !           175: applications include: stock market portfolio management, sales data, 
        !           176: voting patterns, sports (48 statistics on 459 NBA players, in 27 teams, 
        !           177: in four leagues), etc.
        !           178:  
        !           179:  
        !           180: Dynamic queries allow rapid adjustment of query parameters and immediate 
        !           181: display of updated result sets.  These animations enable users to 
        !           182: develop intuitions, discover patterns, spot trends, find exceptions, and 
        !           183: see anomalies.  The Dynamic HomeFinder prototype (Figure 2) allows users 
        !           184: to adjust the cost, number of bedrooms, location, etc. and see points of 
        !           185: light come and go on a map to indicate a matching home.  Users execute
        !           186: up to 100 queries/second (rather than one query per 100 seconds)
        !           187: producing a revealing animated view of where high or low priced homes
        !           188: are found, and there are no syntax errors.  Clicking on a point of
        !           189: light brings up a description or image (videotape available, or for 
        !           190: an empirical comparison with a natural language system, see 
        !           191: Williamson, C. and Shneiderman, B., The Dynamic HomeFinder: Evaluating 
        !           192: dynamic queries in a real-estate information exploration system, 1992 
        !           193: ACM SIGIR Proceedings).  
        !           194: 
        !           195: Dynamic queries are very effective when a visual environment such as a
        !           196: map, calendar, or schematic diagram are available, but they can be
        !           197: easily applied with standard text file output (Figure 3).  Dynamic 
        !           198: queries exemplify the future of interaction; You don't need to
        !           199: describe your goals, negotiate with an intelligent agent, and wait for
        !           200: a response, you Just Do It!   Furthermore, dynamically seeing the 
        !           201: results enables you to explore and rapidly reformulate your goals in 
        !           202: an engaging videogame-like manner.  
        !           203:  
        !           204:  
        !           205: Open problems in information visualization include screen organization, 
        !           206: widget design, algorithms for rapid search and display, use of color
        !           207: and sound, and strategies to accommodate human perceptual skills.  
        !           208: We also see promise in expanding macro makers into the graphical 
        !           209: environment with visual triggers based on controlled replay of 
        !           210: desired actions - the 
        !           211: general idea is Programming in the User Interface (PITUI) to 
        !           212: Do-What-I-Did (DWID).
        !           213:  
        !           214:  
        !           215: I want to encourage the exploration of new metaphors and visions of how 
        !           216: computers can empower people by presenting information, allowing rapid 
        !           217: selection, supporting personally specified automation, and providing 
        !           218: relevant feedback.  Metaphors related to controlling tools or machines 
        !           219: such as driving, steering, flying, directing, conducting, piloting, 
        !           220: or operating seem more generative of effective and acceptable 
        !           221: interfaces, than 'intelligent' machines.
        !           222: 
        !           223: 
        !           224: A scientific approach to user interface research
        !           225:  
        !           226: Whether you agree with the design philosophy in this essay, and
        !           227: especially if you disagree, I hope that you will add to our scientific
        !           228: knowledge by conducting well-designed empirical studies of learning
        !           229: time, measuring performance time for appropriate tasks, recording error
        !           230: rates, evaluating human retention of interface features, and assessing
        !           231: subjective satisfaction.  There's much work to be done to make
        !           232: computing accessible, effective, and enjoyable.
        !           233:  
        !           234:  
        !           235: Acknowledgements:  This essay was prompted by the discussion between 
        !           236: Mark Weiser and Bill Hefley, stimulated by lively email and personal 
        !           237: discussions with Paul Resnick, Tom Malone, and Christopher Fry at MIT, 
        !           238: and refined by comments from Catherine Plaisant, Rick Chimera, Brian 
        !           239: Johnson, David Turo, Richard Huddleston, and Richard Potter at the 
        !           240: Human-Computer Interaction Lab at Univ. of Maryland.  I appreciate Bill 
        !           241: Curtis's support for this vision.  Thanks to all.

This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.