# 21.10.19 - Skype with Luke Church (10:50 - 12:00)

## Introduction
- Names
- showing wiki
  - research
  - scratch-files

how do you like lively so far?
good libraries, but also good to know how to work without

Africa's Voices
core principles: humanitarian work done to the people, not with them -> doesn't work very well
people come from western Europe, life experience is different, difficult to design systems
AV -> system to learn from people, learning what their lives are like

traditional method: survey, people walking around asking people, fly back to work country, analysis, report 
difficulties: walking around the countries safely, understanding the problem well enough to develop options (list needs to be complete), biased facts, getting "wrong" answers (trying to get rid of people or satisfy them)

socio-technical: radioshows, interviews, asking people questions, people message back with open text, getting different views, perspective, wider set of opinions, more complicated
try to turn that into a result to show
people texts and messages, example messages, tool "coding schemes" categories, producing graphs from categories
testing whether different groups of people have different opinions
don't just report results, plus quotes from real citizens, politicians respond differently, more personal
that's what the person said, this is how we translated it, this is the category we put it in
export as csv for excel visualization
idea: zooming in, showing real examples of what people were saying, see what groups of people answered what (geographically, status)
example gender bias on pollution of rivers

"blockchain" over collected data (JSON?, Metadata, annotate data with decisions and who made them)
we can work directly with that data structure, provides provenance, but we can also develop our own data structure
data "cleaned", no telephone numbers or other personal data

co-design workshops with people who do policies with data, PowerPoint deck, answer questions, basically to show and explain 
- Microsoft, Power Business Intelligence
- SAP, explore data of a company, going into visualizations
data cubes, olap
qualitative data -> quantitative data, not being able to go back to that decision process
not trying to copy another system

using machine learning on our data?
decision making should always be done by people, mistrust machine learning algorithm's decision
machine learning: grouping, properties of messages, people can decide quickly for larger data set
want people to read and understand the data, so someone is always reading
extension: message looks very different, something surprising, visualization for researches

users: researchers or policy makers
first researchers: easier threshold, can be trained
policy makers: emotional

1. tool for researchers in office
2. tool for researchers for presenting in front of policy makers
3. tool sharing with people who they got info from (general public)

context information: asking back via text (gender, age, more questions/details)
depends on investigation questions

how many people respond: depending on project e.g 30.000 people, 6.000 people, trying to reach 300.000 people over 15 months

Luke will be our "user" for now

codaV2 labeling tool for categories

## next steps
Luke will e-mail links, presentation slides
we will need to sign data protection agreement
logistics: e-mail group, every couple of weeks at the beginning of project
once a week for e-mail, looking at and playing around
e-mail Luke for questions
book recommendations

## further ideas (after skype call)
next questions: why do libraries not suffice? -> text support, drawing on canvas, svg
currently using tableau (Unternehmenssoftware mit Visualisierung)
anfangs erstmal raw JSON
Masterarbeit über Datenbank im Browser -> platzsparender
Software: MIT-Lizenz