SLIDE 1
Exploring the “Magic” of Algorithmic Predictions
Joyce Lee Sejal Popat Soravis Prakkamakul
UC Berkeley School of Information 2019 MIMS Final Presentation
Exploring the Magic of Algorithmic Predictions Joyce Lee Sejal - - PowerPoint PPT Presentation
Exploring the Magic of Algorithmic Predictions Joyce Lee Sejal - - PowerPoint PPT Presentation
Exploring the Magic of Algorithmic Predictions Joyce Lee Sejal Popat Soravis Prakkamakul UC Berkeley School of Information 2019 MIMS Final Presentation Why do we think of technology as magic? "Any sufficiently advanced
SLIDE 2
SLIDE 3
"Any sufficiently advanced technology is indistinguishable from magic."
Clarke, A. (1999) [1st pub. 1962, rev. 1973, 1984, 1999]. Profiles of the Future: An Inquiry Into the Limits of the Possible. New York, NY: Holt, Rinehart & Wilson.
SLIDE 4
Clarke, A. (1999) [1st pub. 1962, rev. 1973, 1984, 1999]. Profiles of the Future: An Inquiry Into the Limits of the Possible. New York, NY: Holt, Rinehart & Wilson.
"Any sufficiently advanced technology is indistinguishable from magic."
SLIDE 5
What underlies algorithmic “magic”?
1
Proprietary concerns
2
Technical illiteracy
Burrell, J. (2016). How the Machine ‘Thinks’: Understanding Opacity in Machine Learning Algorithms. Big Data & Society. https://doi.org/10.1177/2053951715622512.
SLIDE 6
Astrology Aura Reading Mediumship Numerology Palm Reading Physiognomy Phrenology Psychometry Tarot Card Reading
But attempting to predict is nothing new…
SLIDE 7
Objective
Can reframing algorithmic predictions with the deliberately mystical lens of divination lead people to reconsider its sense of “magic”?
SLIDE 8
Approach
Within the traditions of reflective & speculative design, we develop an interactive installation that enables technology-mediated tarot-card readings.
SLIDE 9
Research Questions
1
Will people question how algorithmic prediction works in the context of divination?
2
Will people resist sharing personal data for the purpose
- f “algorithmic divination”?
SLIDE 10
System Design
SLIDE 11
SLIDE 12
Interaction flow
Conversational Interface Text Analysis & Visualization Personality Prediction Tarot Card Reading
SLIDE 13
Topic Visualization Interior Lamp Black Curtains Card Dispensing Slot
SLIDE 14
Tarot Card Printer Monitor Laptop Microphone Array
SLIDE 15
Tarot Card Printer Monitor Laptop Microphone Array
SLIDE 16
SLIDE 17
Tarot Card Printer Monitor Laptop Microphone Array
SLIDE 18
Tarot Card Printer Monitor Laptop Microphone Array
SLIDE 19
Future Prediction Past Prediction Present Prediction
SLIDE 20
Raw Speech Speech Recognition Intent Recognition Topic Analysis (Empath) Sentiment Analysis MBTI Personality Analysis (SpaCy) Future Prediction Past Prediction Present Prediction Basic Interactions: Skip, Repeat, etc.
SLIDE 21
Raw Speech Speech Recognition Intent Recognition Topic Analysis (Empath) Sentiment Analysis MBTI Personality Analysis (SpaCy) Future Prediction Past Prediction Present Prediction Basic Interactions: Skip, Repeat, etc.
SLIDE 22
Raw Speech Speech Recognition Intent Recognition Topic Analysis (Empath) Sentiment Analysis MBTI Personality Analysis (SpaCy) Future Prediction Past Prediction Present Prediction Basic Interactions: Skip, Repeat, etc.
SLIDE 23
Tarot Card Printer Monitor Laptop Microphone Array
SLIDE 24
Tarot Card Printer Monitor Laptop Microphone Array
SLIDE 25
SLIDE 26
Research Approach
SLIDE 27
10 Participants
Ages 25-40 50% female
Sample Recruited via Email Self-Guided Tarot Reading Follow-Up Interview
SLIDE 28
Resulting Themes
1
Questioning whether or not to share personal data
2
Interrogating underlying technology
SLIDE 29
Resulting Themes
1
Questioning whether or not to share personal data
2
Interrogating underlying technology
SLIDE 30
Difficult questions “The questions got . . . pretty personal. ”
SLIDE 31
Presence of a camera “I remembered that it was being recorded, like there’s a video. So I wondered, ‘Should I be giving all this information out? ”
SLIDE 32
Speaking vs. writing “There is a sense in opening your own mouth to say things that you’re handing over more information… There’s something about talking that makes it feel different”
SLIDE 33
Resulting Themes
1
Questioning whether or not to share personal data
2
Interrogating underlying technology
SLIDE 34
Resulting Themes
1
Questioning whether or not to share personal data
2
Interrogating underlying technology
SLIDE 35
Reflection vs. inception “I wasn’t sure whether I was being reflected or lead”
SLIDE 36
SLIDE 37
Contesting classification… “If I see I’m kind of pigeon holed… I either reassess myself, or question what the algorithm is latching onto”
SLIDE 38
… or being unable to exercise agency “There’s no real agency to clarify what you mean”
SLIDE 39
General sense of opacity “Its mysterious. I mean we assume they’re working. A lot of prediction is business oriented and we don’t know what’s being
- maximized. ”
SLIDE 40
Research Questions, Revisited
1
Will people question how algorithmic prediction works in the context of divination?
2
Will people resist sharing personal data for the purpose
- f “algorithmic divination”?
SLIDE 41
Insights
1
Participants were most interested in the process of generating predictions, rather than the
- utcomes themselves
SLIDE 42
Insights
2
Participants felt they had very little agency to control how they were “seen” by the system; many wondered if they were being reflected or led
SLIDE 43
Implications
1
The magical, optimistic presentation of algorithmic predictions may be doing more harm than good
SLIDE 44
Implications
2
We see a need for transparent communication of technology’s inner workings, even if this requires purposeful ambiguity and room for interpretation
SLIDE 45
Why do we think of technology as magic?
SLIDE 46
Many Thanks To . . .
Our advisor, Kimiko Ryokai Our research participants Our I School professors & department staff Our friends & family (that’s you!)
SLIDE 47
Exploring the “Magic” of Algorithmic Predictions
Joyce Lee Sejal Popat Soravis Prakkamakul
UC Berkeley School of Information 2019 MIMS Final Presentation