Skip to content

Latest commit

 

History

History
383 lines (226 loc) · 9.39 KB

dssg-ds-adc.md

File metadata and controls

383 lines (226 loc) · 9.39 KB

slidenumbers: true

autoplay loop mute

Data Science

for

Social Good

The best minds of my generation are thinking about how to make people click ads. That sucks. -- Jeff Hammerbacher (Co-founder Facebook)

^ now cloudera

Social organizations need to allocate scarce resources for maximum impact

  • predicting where services will be needed
  • prioritizing resources based on expected impact
  • forecasting trends and changes
  • identifying supporters likely to donate

[fit] Basically regression problems

^ Given enough data

Social organizations often don’t have

[fit] data scientists

[fit] What if data scientists would

[fit] donate

[fit] their skills?

right

This is already happening...1

A guided tour through the

[fit] Data Science for Social Good

movement

DSSG Fellowship

  • supported by the Eric & Wendy Schmidt Foundation
  • 6 month fellowship
  • currently in 3rd year (started 2013)
  • Partners: NGOs, Governments

http://dssg.io/

^ started by Rayid Ghani (Obama’s Chief data scientist)

Examples (project headline copy&paste)

  • World Bank Group – Prediction & Identification of Collusion in International Development Projects
  • Chicago Public Schools – Student Enrollment Prediction for Budget Allocation
  • Pecan Street , WikiEnergy – Building Open Source Tools to Analyze Smart Meter Data

More examples

visit http://dssg.io/projects/

DSSG Atlanta

  • like DSSG Chicago
  • mainly funded by Oracle and Georgia Tech
  • started 2014 (one year after Chicago)

http://dssg-atl.io/

DSSG Washington

  • University of Washington just announced their DSSG Summer Program

http://escience.washington.edu/what-we-do/data-science-for-social-good

Bayes Impact (YC-backed)

Bayes Impact is a nonprofit that deploys data scientists to solve big social problems with civic and nonprofit organizations

  • founded 2014
  • started as (full-time-)fellowship
  • now hiring long-term employees

Examples (project headline copy&paste)

  • Increasing Graduation Rate And Optimizing Class Offerings For UC Riverside
  • Improving Outcomes For Emotionally And Behaviorally Challenged Children With Youth Villages
  • Stratification Of Parkinson's Disease Patients
  • Optimizing Ambulance Response Times In Sf

Bayes Hack Nov 15th

right fit

One weekend, impact the world

http://bayeshack.challengepost.com/submissions

^ curious to see how many will be alive in 6 months ^ image on right shows winner

drivendata

  • started this year (2014)
  • currently 4 open competitions

http://www.drivendata.org/

^ https://www.kaggle.com/c/kdd-cup-2014-predicting-excitement-at-donors-choose ^ donors choose lets teachers enter projects for crowdfunding

[fit] more opportunities to get involved

http://blog.datalook.io/definitive-guide-data-science-good/

Data__Kind__

We're tackling the world's biggest problems
 through data science. -- http://www.datakind.org

DataKind connects charities with data scientists by organizing two-day data dives where those data scientists help solve the charities’ data problems.

Jake Porway at TEDxMontreal

left

[fit] Examples

Analyzing poverty through satellite imagery

fit right

DataKind helped GiveDirectly – an NGO making unconditional cash transfers to poor households via mobile phones in Kenia and Uganda2 – to identify especially needy villages through satellite image analysis3.

^ predictive model to estimate number of roofs ^ and percentage of thatched / metal roofs ^ crowdsourced training data ^ template matching ^ 100 person days of manual effort saved

[fit] View the presentation

[fit] or read the paper

right

Early warning system for human rights abuses

To help prioritize the many calls for help reaching Amnesty International’s Urgent Action Network DataKind volunteers have created a predictive model that analyzes messages for potential escalation.45

Predicting demand for children’s hospices

left

Combining data from Shooting Star Chase, public data about the hospice and healthcare sector and demographic data DataKind volunteers calculated predicted demand against hospice capacity to reveal areas of possible shortage.6

^ + a few other things

Most of DataKinds projects have been tackled by volunteers on 2-day data dives.

What’s a

[fit] Data Dive?

^ Who has been on a data dive?

Before: Preparation of the data

(by voluntary data ambassadors in collaboration with the challenge partner – starting ~2 month before the data dive)

  • anonymization/pseudonymization
  • cleaning/fixing
  • ensuring proper (machine readable) data formats

Before:

Preparation of the questions

Any data scientist worth their salary will tell you that you should start with a question, NOT the data. -- Jake Porway in https://hbr.org/2013/03/you-cant-just-hack-your-way-to/

The Data Dive

  • Challenge partners pitch their problems
  • Volunteers create analyses, models and visualizations (led by data ambassadors) in two intense days of hacking
  • solutions are being presented at the end

^ Big community event ^ Data Ambassadors important

Aftermath:

Implementation

Social organizations still don’t have the expertise: data ambassadors must help implement the solutions

^ Not yet quite clear to me ^ Sent DataKind and email to clarify

Aftermath:

Impact Analysis

Did it help at all?

^ Not yet quite clear to me

There is currently no organization in Germany comparable to DataKind.

right

There is currently no organization in Germany comparable to DataKind.

100%

100%

We want to start the

[fit] Data Science for Social Good

Movement in Germany

Who we are?

  • Daniel Kirsch
  • Marit Brademann
  • Jana Kludas
  • Richard Lawrence
  • Georg Walther

^ Detexify, Co-founded OK Lab Münster

Supporters

  • Klaas Bollhöfer, Chief Data Scientist @ The Unbelievable Machine Company
  • Adam Drake, Chief Data Officer @ Skyscanner
  • Dr. Alexander Weiß Head of Data Analytics @ Trademob

What we need

We need data ambassadors

(how DataKind calls them)

  • to prepare data before data dives
  • lead teams at data dives
  • help with the implementation afterwards

We need problems

The international of the Data Science for Social Good-movement shows that data scientists are eager to donate their skills.

Social organizations need to understand how we can help them. Are you in contact with NGOs? Spread the word!

Contact

http://dssg-berlin.org/ @dssgber

Daniel Kirsch [email protected]

autoplay loop mute

[fit] Thank you!

[fit] Questions?

Links:

Attribution

Footnotes

  1. ...but not so much in Germany ↩

  2. http://www.ted.com/talks/joy_sun_should_you_donate_differently ↩

  3. http://www.datakind.org/projects/using-the-simple-to-be-radical/ ↩

  4. http://www.datakind.org/projects/using-predictive-analytics-to-prevent-human-rights-abuses/ ↩

  5. http://www.washingtonpost.com/business/on-it/amnesty-international-considers-using-big-data-to-predict-human-rights-violations/2013/11/22/3f4f1a1e-5388-11e3-a7f0-b790929232e1_story.html ↩

  6. http://www.datakind.org/finding-30000-missing-children/ ↩