Information
Designing Machine Learning is a project by the Stanford d.School to make Machine Learning (ML) more accessible to innovators from all disciplines. We believe that ML will soon be a widespread feature of products, services, systems, and experiences in all walks of life. In order to make this computational revolution possible, as well as to ensure that it closely follows human values, we must empower a new generation of professionals to incorporate ML into their creative process.
Quick Navigation
Motivation
Design is the exploration of possibility, at once a critique of the present and a supreme optimism about human ingenuity. In a way, machines are design made manifest. Machines are programmable constructs, giving limitless potential and new forms of agency to the designer. Machine learning is yet another axis of empowerment, still incipient in its capabilities. Perhaps it is the humility with which we have previously approached machine learning that has prevented us from applying it pervasively, or exposing it to wider inquiry.
Traditionally, designers have differed from data scientists in their approach to building systems and their metrics of measuring success. At its core, this is a difference of values that pervades each field. Data science is statistical and numerical, with focus placed on standardized measures, such as “training versus test accuracy”. To designers, this (and many other) metrics are essentially arbitrary. What matters to designers is the human value ascribed to the experience of the product. How do we reconcile this stark contrast?
Course Information
Office Hours
As a courtesy, for appointments please schedule at least four hours in advance to give us time to prepare.
- Abhay’s Office Hours: Abhay’s Scheduling Link
- Michelle’s Office Hours: By appointment via e-mail ([email protected]) and Wednesdays from 2PM-3PM in d.School Bay Studios (please e-mail to confirm space).
- Anand’s Office Hours: By appointment via e-mail ([email protected]) and Thursdays from 5PM-6PM in d.School Bay Studios (please e-mail to confirm space).
Music From Class
Week 1
Topics covered: Intro to Machine Learning, Decision-Making Systems, Class Logistics.
Assignments Due Next Week
- Take the course intro survey
Readings Due Next Week
- Deloitte: State of AI in Enterprise
- Quartz: The Data that Transformed AI Research
- The Verge: Wikipedia Engages Nuclear Option
- [Human-Centered Machine Learning]
Week 2
Topics covered: Understanding the Data Science Pipeline. Data Collection Systems. Content Filtering in Wikipedia
Content From Class
- Data: The Raw Material
- Data Flow in an ML System
- Colab Notebook on Data Science Pipeline
- Content Filtering in Wikipedia
- Colab Notebook on Wikipedia Filtering
Assignments Due Next Week
- Presentations for first guided investigation, Wikipedia Abuse Filter
- Fill out and bring your brainstorming sheets along with print-outs of any materials you will turn in.
Readings Due Next Week
- Recommender Systems: From Algorithm to User Experience
- Algorithms in practice: Comparing web journalism and criminal justice
- The Netflix Prize: How a $1 Million Contest Changed Binge-Watching Forever
Week 3
Topics covered: Machine Learning + User Experience. Supervised Learning. Recommender Systems & Netflix Prize.
Content From Class
Assignments Due Next Week
- Presentations for second guided investigation, Netflix Recommendation System
- Fill out and bring your brainstorming sheets along with print-outs of any materials you will turn in.
Readings Due Next Week
- The Unreasonable Effectiveness of Data
- Are Personas Done?
- A Visual Introduction to Machine Learning. Read parts 1 & 2. Typed notes are okay as a hand-in for this reading.
- Data-Driven Personas at Salesforce
Week 4
Topics covered: Present Netflix Projects. Prototyping in ML. Data Visualization & Data Exploration. Clustering Data. Project on Meetup Data Set.
Content From Class
- ML Basics (updated)
- Data Viz
- Meetup Activity
- Meetup.com API Tableau Visualization
- Meetup.com clustering
- Raw data from Meetup can be found here
- Category Membership Model (Michelle): here is the clustering .ipynb file with annotations, here are the results, and here are the output clusters as .csv
- Group Descriptions Model (Abhay): you can play around with the colab notebook here
Assignments Due Next Week
- None
- Design Project #1 due in Week 6: Design Project: Meetup
Readings For Next Week
- Deep Learning
- Machine Learning Yearning Sections 13 - 19
- Image Style Transfer Using Convolutional Neural Networks OR non-technical intro to same content
Resources Mentioned In Class
Week 5
Topics covered: Introduction to Neural Network Systems and High-Dimensional Classifiers. Image & Video Processing.
Content From Class
Assignments Due Next Week
Readings For Next Week
- Ten Questions Concerning Generative Computer Art
- Teaching Machines to Draw
- StyLit: Illumination-Guided Example-Based Stylization of 3D Renderings
- GANGough: Creating Art with GANs
- AI Enters Dance Dance Revolution
Week 6
Topics covered: Present Meetup Project Work. Continued Study on Neural Networks. Generative Design & GANs.
Content From Class
- Drawing with Neural Networks
- Generative Design
- Chris Donahue: Pairing human control with generative models for creative content synthesis
Assignments Due Next Week
- None
Readings For Next Week
- Analogy as the Core of Cognition
- Better Language Models
- Alexa, Should we Trust You?
- The Siri Experience
- Lilt is building a machine translation business with humans at its core
Week 7
Topics covered: Natural Language Processing and Voice Interfaces. Conversational UI Activity. Second Project Begins with Lilt.
Content From Class
Assignments Due Next Week
- None
- Design Project #2 due in Week 9: Design Project: Lilt
Readings For Next Week
- Engelbart: Augmenting Human Intellect
- ML: The high interest credit card of tech debt
- Power to the People: The Role of Humans in Interactive Machine Learning
- Robert from Lilt discussing OpenAI GPT-2
Week 8
Topics covered: Human-Machine Interaction. Social and Cultural Factors in Machine Learning. Ethics and Bias.
Content From Class
Assignments Due Next Week
- Design Project #2: Design Project: Lilt
- Play around with the “…what if” tool, and take note of any findings.
Readings For Next Week
- How to Prevent Discriminatory Outcomes in Machine Learning
- How AI Changed Organ Donation in the US
- The Moral Machine Experiment
Week 9
Topics covered: Present Second Project Work. Human Factors in Designing ML. Continued Work on Ethics.
Speakers From Class
- Elger Oberwelz, Executive Design Director at IDEO
- Christian Ramsey, Data Scientist and Researcher at IDEO
- Andrea Gagliano, Lead Visual Intelligence Researcher at Getty Images
Readings For Next Week
- None
Week 10
Topics covered: Disruption and innovation in ML, Consequences of ML Systems.
Content From Class
- Pre-Mortem Activity
- Disruption and Innovation
- Please fill out the anonymous survey form here to provide us feedback on the course.