Top Banner
Eric Horvitz Eric Horvitz Microsoft Research & Microsoft Research & University of Washington University of Washington Toward Models of Toward Models of Surprise Surprise
49

Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Mar 26, 2015

Download

Documents

James Mitchell
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Eric HorvitzEric Horvitz

Microsoft Research & Microsoft Research &

University of WashingtonUniversity of Washington

Toward Models of SurpriseToward Models of Surprise

Page 2: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

SurpriseSurprise

Unexpected eventUnexpected event: “: “We thought about it, but We thought about it, but considered it would occur with probability < xconsidered it would occur with probability < x””

Unmodeled eventUnmodeled event: : “We never even thought about “We never even thought about that…”that…”

Significant consequenceSignificant consequence: : Significant positive or Significant positive or negative swing in utility of outcome for observer or negative swing in utility of outcome for observer or groupgroup

Events can be personal, local, national, worldEvents can be personal, local, national, world

Unexpected Unexpected or or umodeledumodeled future event or future event or outcome of outcome of significant consequencesignificant consequence..

Page 3: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

SurpriseSurprise

Surprise defined in terms of an observer’s Surprise defined in terms of an observer’s expectationsexpectations

Deliberation and modeling can influence degree of Deliberation and modeling can influence degree of or nature of surpriseor nature of surprise Moving from surprise to plausible expectationMoving from surprise to plausible expectation

Expectation of classes or properties of surprise should Expectation of classes or properties of surprise should

given that a surprise occurs.given that a surprise occurs.

Unexpected Unexpected or or umodeledumodeled future event or future event or outcome of outcome of significant consequencesignificant consequence..

Page 4: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Forecasting SurpriseForecasting Surprise

Qualitative brainstorming about likelihood and nature of Qualitative brainstorming about likelihood and nature of surprises over future time periods from knowledge, trends, surprises over future time periods from knowledge, trends, and historical dataand historical data

Opportunity: Statistical models of surpriseOpportunity: Statistical models of surprise Learn from databases of past surprise to predict future surprisesLearn from databases of past surprise to predict future surprises

Opportunity: Statistical models of predictive competencyOpportunity: Statistical models of predictive competency Context- and content-centric failure to predict with accuracyContext- and content-centric failure to predict with accuracy

Qualitative and quantative approachesQualitative and quantative approaches

Page 5: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Last 12 monthsLast 12 monthsTop Top nn PropertiesProperties

SurpriseCasting: Abstract and ProjectSurpriseCasting: Abstract and Project Example:Example:

Identify top Identify top nn most surprising things in prior 1 year (world most surprising things in prior 1 year (world

events or personal life)events or personal life)

Repeat for 5 years of surprisesRepeat for 5 years of surprises

Characterize nature and rate of significant surprisesCharacterize nature and rate of significant surprises

Consider top Consider top nn over longer-term periods in same manner over longer-term periods in same manner

20042004 20032003 20022002Top Top nn PropertiesProperties Top Top nn PropertiesProperties Top Top nn PropertiesProperties

Page 6: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Surprise Forecasting: 2005-2020Surprise Forecasting: 2005-2020

Surprise TypesSurprise Types Examples Examples

Attention & Information Attention & Information

ActionsActions

SurpriseCasting: Abstract and ProjectSurpriseCasting: Abstract and Project

Top Top nn surprise surprise types, by class types, by class or propertiesor properties

Assume type Assume type occurs; give top occurs; give top m m examples per typeexamples per type

Recommended triage of focus of Recommended triage of focus of attention: additional study, info. attention: additional study, info. gathering, monitoring per typegathering, monitoring per type

Recommended actions to Recommended actions to react, contain, disrupt react, contain, disrupt negative suprises and/or negative suprises and/or exploit positive surprisesexploit positive surprises

Turn to the future with statistics about types, examples. Turn to the future with statistics about types, examples.

Consider triaging of attention and recommending actions.Consider triaging of attention and recommending actions.

Surprises in technical sophistication of weaponry available inexpensively and universally

Largescale simultaneous & continuous disabling of communication and sigint satellites available to multiple states.

Page 7: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Surprise Forecasting: 2005-2020Surprise Forecasting: 2005-2020

Surprise TypesSurprise Types Examples Examples

Attention & Information Attention & Information

ActionsActions

SurpriseCasting: Abstract and ProjectSurpriseCasting: Abstract and Project

Top Top nn surprise surprise types, by class types, by class or propertiesor properties

Assume type Assume type occurs; give top occurs; give top m m examples per typeexamples per type

Recommended triage of focus of Recommended triage of focus of attention: additional study, info. attention: additional study, info. gathering, monitoring per typegathering, monitoring per type

Recommended actions to Recommended actions to react, contain, disrupt react, contain, disrupt negative suprises and/or negative suprises and/or exploit positive surprisesexploit positive surprises

Deliberation about potential surprises: Deliberation about potential surprises:

“Backcasting”“Backcasting” Chaining from surprising outcome to explore feasibilityChaining from surprising outcome to explore feasibility

Exotic transnational actors with access to increasing technical prowess.

Terrorist organization establishes secret alliance with seemingly friendly former adversary state, intolerant of US “empire”; gains access to nuclear arms with deniability for former adversary.

Page 8: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Expecting SurprisesExpecting Surprises

Technical developmentsTechnical developments

International political terrainInternational political terrain

Page 9: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Technical Surprises from PastTechnical Surprises from Past

Big surprises over 50 yearsBig surprises over 50 years

FlightFlight From likely unachievable dream, to major pillar of force From likely unachievable dream, to major pillar of force

projection and deterrence in just a “few summers”…projection and deterrence in just a “few summers”…

Quickly melted into background of obviousnessQuickly melted into background of obviousness In a few years: From awe of a “canvas flying” contraption to In a few years: From awe of a “canvas flying” contraption to

concerns about low-salt meal while streaking across ocean in concerns about low-salt meal while streaking across ocean in widebody jet.widebody jet.

Nuclear fission and fusionNuclear fission and fusion Theory of Relativity and implications out of the blueTheory of Relativity and implications out of the blue

Shifts in balance of powerShifts in balance of power

Risk to nation, human civilizationRisk to nation, human civilization

Page 10: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Classes of Future Technical Surprises?Classes of Future Technical Surprises? ComputationComputation

Intelligent systems change world (consumer, defense, Intelligent systems change world (consumer, defense,

communications, etc.) in a fundamental way?communications, etc.) in a fundamental way?

EnergyEnergy New surprise source of energy changes economics of worldNew surprise source of energy changes economics of world

Changes nature of powerplants in our and adversaries weapon Changes nature of powerplants in our and adversaries weapon

systemssystems

Shift in balance of power hinge on the new sourceShift in balance of power hinge on the new source

Asymmetric force have new tool on their side for unleashing chaos Asymmetric force have new tool on their side for unleashing chaos

and disruptionand disruption

BiologyBiology Assembly of Polio virus via internet and mail order in 2002 heralds Assembly of Polio virus via internet and mail order in 2002 heralds

era of bio-hackingera of bio-hacking

Major disruptions via pandemicsMajor disruptions via pandemics

Page 11: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 12: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Futurist Community: “Wildcards”Futurist Community: “Wildcards”

“A wild card is a future development or event with a relatively low probability of occurrence but a likely high impact on the conduct of business”

BIPE Conseil, Copenhagen Institute for Futures Studies, Institute for the Future: Wild Cards: A Multinational Perspective, Institute for the Future 1992

Page 13: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Futurist Community: “Wildcards”Futurist Community: “Wildcards”

Futurist discussion of wildcards, “futurequakes”Futurist discussion of wildcards, “futurequakes” Are natural, human-caused, or combinationsAre natural, human-caused, or combinations

Are negative and positiveAre negative and positive

Multiple wildcards can interactMultiple wildcards can interact

Breaks, discontinuities vs. trendsBreaks, discontinuities vs. trends

Peterson (2000),Peterson (2000), Out of the Blue: How to Anticipate Big Out of the Blue: How to Anticipate Big

Future SurprisesFuture Surprises

Cornish (2003), Cornish (2003), The Wild Cards in Our FutureThe Wild Cards in Our Future

Steinmüller (2003), The Future as Wild Card: A Short Introduction to a New Concept

Page 14: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

On Expected Surprises…On Expected Surprises…

Earth & SkyEarth & Sky

Biomedical DevelopmentsBiomedical Developments

Geopolitical & Sociological changesGeopolitical & Sociological changes

Technology & Infrastructure UpheavalTechnology & Infrastructure Upheaval

Surprise AttackSurprise Attack

Spiritual & ParanormalSpiritual & Paranormal

J. Petersen, J. Petersen, Out of the Blue: How to Anticipate Big Future SurprisesOut of the Blue: How to Anticipate Big Future Surprises, , 2000. 2000.

Page 15: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

On Expected Surprises…On Expected Surprises…

We are now living in another period of significant We are now living in another period of significant transition--a foreshortened span of time, during which our transition--a foreshortened span of time, during which our surroundings and experiences will change more than surroundings and experiences will change more than during any era in history. during any era in history.

Humanity has never lived through the convergence--and, Humanity has never lived through the convergence--and, in some cases, the collision--of global forces of such in some cases, the collision--of global forces of such magnitude and diversity. magnitude and diversity.

… … One foreseeable outcome might be global instability; One foreseeable outcome might be global instability; another, a planetary renaissance. another, a planetary renaissance.

J. Petersen, J. Petersen, Out of the Blue: How to Anticipate Big Future SurprisesOut of the Blue: How to Anticipate Big Future Surprises, , 2000. 2000.

Page 16: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

On Expected Surprises…On Expected Surprises…

… … In any case, during the next two decades, almost In any case, during the next two decades, almost every aspect of life will be fundamentally reshaped. every aspect of life will be fundamentally reshaped.

. .

J. Petersen, J. Petersen, Out of the Blue: How to Anticipate Big Future SurprisesOut of the Blue: How to Anticipate Big Future Surprises, , 2000. 2000.

Page 17: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Technological Prowess and WildcardsTechnological Prowess and Wildcards

Rapidly growing technological prowess of humans has, Rapidly growing technological prowess of humans has, for the first time in history, produced new classes of wild for the first time in history, produced new classes of wild cards that could potentially destroy the whole human cards that could potentially destroy the whole human race. race.

The new kinds of wild cards are simply too destructive to The new kinds of wild cards are simply too destructive to allow to happen and therefore require active attempts to allow to happen and therefore require active attempts to anticipate them and be proactive in dealing with their anticipate them and be proactive in dealing with their early indicators.early indicators.

""Peterson,Peterson, Out of the Blue: How to Anticipate Big Future Surprises Out of the Blue: How to Anticipate Big Future Surprises, 2000, 2000

Page 18: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Preparing for WildcardsPreparing for Wildcards

Wild cards can be anticipated and prepared for. Wild cards can be anticipated and prepared for.

Thinking about a wild card before it happens is important Thinking about a wild card before it happens is important and valuable. and valuable.

The more that is known about a potential event, the less The more that is known about a potential event, the less threatening it becomes and the more obvious the threatening it becomes and the more obvious the solutions seem.solutions seem.

""Peterson,Peterson, Out of the Blue: How to Anticipate Big Future Surprises Out of the Blue: How to Anticipate Big Future Surprises, 2000, 2000

Page 19: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

BackcastingBackcasting

Cornish: Multipronged Aerial Terrorist Attack was identified,Cornish: Multipronged Aerial Terrorist Attack was identified,

e.g.,1987 article (B. Jenkins); 1994 article (M. Cetron):e.g.,1987 article (B. Jenkins); 1994 article (M. Cetron):

"Targets such as the World Trade Center not only provide the "Targets such as the World Trade Center not only provide the requisite casualties but, because of their symbolic nature, provide more requisite casualties but, because of their symbolic nature, provide more bang for the buck.”bang for the buck.”

"In order to maximize their odds for success, terrorist groups will "In order to maximize their odds for success, terrorist groups will likely consider mounting multiple, simultaneous operations with the aim likely consider mounting multiple, simultaneous operations with the aim of overtaxing a government's ability to respond, as well as of overtaxing a government's ability to respond, as well as demonstrating their professionalism and reach."demonstrating their professionalism and reach."

“ “Despite all this, terrorism will remain a back-burner issue for Despite all this, terrorism will remain a back-burner issue for Western leaders as long as the violence strikes in distant lands and has Western leaders as long as the violence strikes in distant lands and has little impact on their fortunes or those of their constituents.”little impact on their fortunes or those of their constituents.”

E. Cornish,E. Cornish, The Wild Cards in Our Future.The Wild Cards in Our Future. The Futurist. Washington: Jul/Aug  The Futurist. Washington: Jul/Aug 2003.Vol. 37, Iss. 4.2003.Vol. 37, Iss. 4.

Page 20: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

BackcastingBackcastingHow might we proceed to evaluate this warning?How might we proceed to evaluate this warning? Poll experts on likelihood and targets of aerial suicide attacks. Poll experts on likelihood and targets of aerial suicide attacks.

Assume aerial attacks plausible.Assume aerial attacks plausible. Backcasting: Reason about the challenges and solutions.Backcasting: Reason about the challenges and solutions.

Terrorists are unlikely to have their own aircraft. Could they buy one? Perhaps, but the cost would be enormous and authorities might ask too many questions. Could they hijack a military bomber and load it with bombs? They would have to be awfully clever and lucky to get onto an air base, seize a plane, load it with bombs, and fly it to a target. Success would be unlikely, and terrorists would know it.Could they hijack a military bomber and load it with bombs? Could they steal a plane? An individual might have difficulty intimidating passengers and crew, but a group could probably do it even with modest weaponry. Getting into the pilot's cabin seems possible. So taking command of the plane seems feasible. But would the pilot follow instructions? Some instructions, maybe, but perhaps not all-especially if asked to do something suicidal. Could a terrorist be trained as a pilot to take over? Yes; a number of men in terrorist organizations have been trained as pilots.But how can an aircraft cause destruction if it has no bombs to drop? It can simply crash into its target. Such a crash should wreck a building; a skyscraper would be especially vulnerable. Of course, everybody on the plane would be killed, but if the terrorists are ready to die themselves, the deaths of the passengers serve as an added benefit. Scores of terrorists have gone on similar suicide missions by strapping bombs to themselves.

We can now envision a group of terrorists hijacking an airliner and crashing it into an enemy target. But what target might it be? If we had read Cetron's article, we would already have one target to think about-the World Trade Center. In fact, Islamic terrorists did attack the Center in 1993, but their car bomb failed to destroy the building.However, Cetron suggested that multiple targets might be selected for an attack. So we might stress the possibility of a larger, wider attack involving more terrorists and multiple planes and targets.

Page 21: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Research on Research on Statistical Models of SurpriseStatistical Models of Surprise

Most reasoning about surprise qualitativeMost reasoning about surprise qualitative

Challenge: Data sparcityChallenge: Data sparcity

Pushing on theoretical foundations of Pushing on theoretical foundations of surprisesurprise

Example: Research traffic prediction Example: Research traffic prediction service, Seattleservice, Seattle

Page 22: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Streaming IntelligenceStreaming Intelligence

Learning & Learning & reasoningreasoning

Web serviceWeb service

SensingSensing

Rich sensingRich sensingPortablePortabledevicedevice

Continual sensing, learning, and reasoning Continual sensing, learning, and reasoning always available in stream of daily lifealways available in stream of daily life

Page 23: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Predictions in Your PocketPredictions in Your Pocket

Time untiljam

Times until jams clear

Page 24: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

• Event store• Learning• Reasoning

Multiple views on traffic

Operator ID: NickHeading: INCIDENTMessage: INCIDENT INFORMATIONCleared 1637: I-405 SBJS I-90 ACC BLK RL CCTV1623 – WSP, FIR ON SCENE

Incident reportsIncident reports

WeatherWeather

Major eventsMajor events

Traffic Prediction ChallengeTraffic Prediction Challenge

Page 25: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

From Incident Reports to SalienceFrom Incident Reports to Salience e.g., Accidentse.g., Accidents

How many lanes?How many lanes?Emergency vehicles?Emergency vehicles?

Operator ID: Nick Heading: INCIDENT Message:INCIDENT INFORMATION Cleared 1637: I-405 SB JS I-90 ACC BLK RL CCTV 1623 – WSP, FIR ON SCENE

Page 26: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Data Set: Data AbstractionData Set: Data Abstraction

Lane Lane StationStation Bottleneck Bottleneck

Page 27: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 28: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 29: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 30: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 31: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 32: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Building Predictive Models from DataBuilding Predictive Models from Data

Max likely time until bottleneck evaporates

Database of sensed events

• Data store

• System-wide status & dynamics• Incident reports• Sporting events• Weather• Time of day• Day of week• Season• Holiday status

• Base-level predictions

Machine Learning

Bayesian network forecasting model

Page 33: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Predict Future SurprisesPredict Future Surprises

Surprise at time Surprise at time tt in thein the futurefuture: Infer surprise with : Infer surprise with model built from training setmodel built from training set

EvidenceEvidenceTime of day, day of weekTime of day, day of week

Bottlenecks, flows and their durationsBottlenecks, flows and their durations

Incident reportsIncident reportsWeatherWeather

GamesGames

AccidentsAccidents

Page 34: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Building Models of SurpriseBuilding Models of Surprise

Learning predictive model of surpriseLearning predictive model of surprise Model of user expectation and supriseModel of user expectation and suprise Models of future surprisesModels of future surprises

Expectations

Human Forecaster

• • Major events• Weather• Time of day• Day of week• Holiday status

RealWorld

Outcome

• System-wide status & dynamics• Incident reports• Sporting events• Weather• Time of day• Day of week• Season• Holiday status

Database of surprising events

• Data store

Future traffic Future traffic

Prob

abilityP

robability

!!

Events at t

Page 35: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Building Models of Building Models of FutureFuture Surprises Surprises

Learning predictive model of surpriseLearning predictive model of surprise Model of user expectation and supriseModel of user expectation and suprise Models of future surprisesModels of future surprises

Expectations

Human Forecaster

• • Major events• Weather• Time of day• Day of week• Holiday status

RealWorld

Outcome

Database of surprising events

• Data store

• System-wide status & dynamics• Incident reports• Sporting events• Weather• Time of day• Day of week• Season• Holiday status

Future traffic Future traffic

Prob

abilityP

robability

!!

Events at t

Machine learning

• System-wide status & dynamics• Incident reports• Sporting events• Weather• Time of day• Day of week• Season• Holiday status

Events at t-T

Page 36: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Traffic forecasting service

• Data store• Inference• User logs

• System-wide status & dynamics• Incident reports• Sporting events• Weather• Time of day• Day of week• Season• Holiday status

• Base-level predictions

• Context-sensitive error models

Adding Surprise ModelingAdding Surprise Modeling

Surprise & surprise forecasting

• Surprise forecasting models

Competencyannotation

Page 37: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Where will there be surprises in time Where will there be surprises in time tt??

Page 38: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Sample Models and ExplorationSample Models and Exploration Influence of Seattle sporting eventsInfluence of Seattle sporting events

Page 39: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Sample Models and ExplorationSample Models and Exploration Influence of weatherInfluence of weather

Page 40: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Sample Models and ExplorationSample Models and Exploration Influence of an accident at Bottleneck 15Influence of an accident at Bottleneck 15

Page 41: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Influence of accidents at B15 and B3 on anomaly at B4 at Influence of accidents at B15 and B3 on anomaly at B4 at 30 minutes30 minutes

Examining DetailsExamining DetailsAnomaly at B4Anomaly at B4Acc 15 Acc 15 .5 surprise .5 surprise -low-low

Acc 3 Acc 3 .5 surprise .5 surprise-.25 low-.25 low-.25 high-.25 high

Page 42: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 43: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Alerting on Device and DesktopAlerting on Device and Desktop

Specify preferences and download to Specify preferences and download to devicedevice

Time- and route-specific preferencesTime- and route-specific preferences Routes composed from bottlenecksRoutes composed from bottlenecks Multiple alert typesMultiple alert types

Page 44: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 45: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 46: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.
Page 47: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Models of SurpriseModels of Surprise

Definitions of surpriseDefinitions of surprise

Qualitative and quantitativeQualitative and quantitative

Abstraction about past surprises to Abstraction about past surprises to expect classes of surpriseexpect classes of surprise

Wild card notion in world of futuristsWild card notion in world of futurists

Directions: Statistical models of surpriseDirections: Statistical models of surprise

Page 48: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.

Surprise Forecasting: 2005-2020Surprise Forecasting: 2005-2020

Surprise TypesSurprise Types Examples Examples

Attention & Information Attention & Information

ActionsActions

SurpriseCastSurpriseCast

Top Top nn surprise surprise types, by class types, by class or propertiesor properties

Assume type Assume type occurs; give top occurs; give top m m examples per typeexamples per type

Recommended triage of focus of Recommended triage of focus of attention: additional study, info. attention: additional study, info. gathering, monitoring per typegathering, monitoring per type

Recommended actions to Recommended actions to react, contain, disrupt react, contain, disrupt negative suprises and/or negative suprises and/or exploit positive surprisesexploit positive surprises

Page 49: Eric Horvitz Microsoft Research & University of Washington Toward Models of Surprise.