Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
1
AI for Social Good: Key Techniques, Applications, and Results
AI for Social Good: Key Techniques, Applications, and Results
MILIND TAMBE
Founding Co-director, Center for Artificial Intelligence in Society (CAIS)
University of Southern California
Co-Founder, Avata Intelligence
USC Center for Artificial Intelligence in Society (CAIS.USC.EDU)
6/27/2017 2
2
Mission Statement: Advancing AI research driven by…
Grand Challenges of Social Work
Ensure healthy development for all youth
Close the health gap
Stop family violence
Advance long and productive lives
End homelessness
Achieve equal opportunity and justice
…
6/27/2017 3
Overview of CAIS Project Areas
AI for Assisting Low Resource Communities
Social networks: Spread HIV information, influence maximization
Real-world pilot tests: Big improvements
6/27/2017 4
3
Overview of CAIS Project Areas
6/27/2017 5
Machine learning/planning: Predicting poaching spots, patrols
Real-world: Uganda, South Asia…
AI for Earth
Game theory: security resource optimization
Real-world: US Coast Guard, US Federal Air Marshals Service…
Overview of CAIS Project Areas
6/27/2017 6
AI for Public Safety and Security
4
AI Program: HEALER
6/27/2017 7
HIV Prevention Programs:Using Social Networks to Spread HIV Information
6/27/2017 8
5
Challenge: Adaptive selection in Uncertain Network
K = 51st time step
6/27/2017 9
106/27/2017
6
Challenge: Adaptive selection in Uncertain Network
K = 52nd time step
6/27/2017 11
Challenge 3 : Adaptive selection
K = 53rd time step
6/27/2017 12
NO LONGER A SINGLE SHOT DECISION PROBLEM
7
6/27/2017 13
Creating an Adaptive Policy: “POMDP” [2015]
Homeless shelters – sequentially select nodes under uncertainty
Policy driven by observations about edges
Action
Choose youth forintervention
Observation: Which edges exist?
Adaptive Policy
Algorithm to FINDADAPTIVE POLICY
Actual socialNetwork in the
real world:
Real Networks – Simulation Results [2016-2017]
6/27/2017 14
0
10
20
30
40
50
60
70
80
90
Venice Hollywood
Indirect Influen
ce
Degree CR PSINET HEALER‐1 HEALER‐2
8
Pilot Testswith 170 Homeless Youth [2017]
Recruited youths:
Preliminary network —> HEALER
Bring 4 youth for training, get edge data —> HEALER
Bring 4 youth for training, get edge data —> HEALER
Bring 4 youth for training
HEALER HEALER++ DEGREE CENTRALITY
62 56 55
6/27/2017 15
Results: Pilot Studies
6/27/2017 16
0
20
40
60
80
100
HEALER HEALER++ Degree
Percent of non-Peer Leaders
Informed Not Informed
0
20
40
60
80
100
HEALER HEALER++ Degree
Informed Non‐Peer Leaders Who Started Testing for HIV
Testing Non‐Testing
9
AI Program: HEALER
6/27/2017 17
Next Steps
900 youth study begun at three locations in Los Angeles
300 enrolled in HEALER/HEALER++
300 enrolled in no condition
300 in Degree centrality
6/27/2017 18
“Picking youth as peer leaders waschanging their self esteem and thesense of confidence that they couldbe an agent for positive change….”
Eric Rice
10
Overview of CAIS Project Areas
AI for Assisting Low Resource Communities
Substance abuse, suicide prevention…
Modeling gang violence, matching homeless and homes…
6/27/2017 19
Protecting Wildlife in Uganda
6/27/2017 20
11
Predicting Poaching from Past Crime Data
PAWS: Applying AI for protecting wildlife
Poacher Behavior Prediction
6/27/2017 21
Data from Queen Elizabeth National Park, Uganda
Poacher behavior prediction [2016]
Number of poaching attacks over 12 years: ~1000
How likely is an attack on
a grid Square
Ranger patrolfrequency
Animal density
Distance to rivers / roads
Area habitat
Area slope
…
6/27/2017 22
12
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
L&L
Uniform Random SVM CAPTURE Decision Tree Our Best Model
Poacher Behavior Prediction
Poacher Attack Prediction [2017]
Results from 2015
6/27/2017 23
6/27/2017 24
Real-world Deployment: (1 month)
13
Real-world Deployment: Results
Two 9 sq KM patrol areas: Predicted hot spots with infrequent patrols
Poached Animals: Poached elephant Snaring: 1 elephant snare roll Snaring: 10 Antelope snares
Historical Base Hit Rate
Our Hit Rate
Average: 0.73 3
6/27/2017 25
6/27/2017 26
0
0.02
0.04
0.06
0.08
0.1
0.12
High (1) Low (2)
Catch per Unit Effort
Experiment Group
Real-world Deployment: Field Test 2 (6 months) [2017]
Catch Per Unit Effort (CPUE)
Unit Effort = km walkedHistorical CPUE: 0.04Our high CPUE: 0.11
14
6/27/2017 27
Field Test Side Effects:Queen Elizabeth National Park
Rangers followed poachers’ trail; ambushed camp
Arrested one (of 7) poachers
Confiscated 10 wire snares, cooking pot, hippo meat, timber harvesting tools.
Pursuit of poachers
Signs of road building, fires, illegal fishing
AI for Social Good
6/27/2017 28
15
THANK YOU
CAIS.USC.EDU
THANK YOU
CAIS.USC.EDU
Towards the Future: AI for Social Good
Significant potential: AI for low resource communities, emerging markets
Not just applications; novel research challenges: Fundamental computational challenges from use-inspired research Designing AI systems in society:
• Interpretability• Complementing human autonomy
Methodological challenges: Encourage interdisciplinary research: measures impact in society
6/27/2017 30