Introduction to Social Data Science Session 1 – Introduction to Social Data Science

class: center, middle, inverse, title-slide

.title[
# Introduction to Social Data Science<br><span class="session">Session 1 – Introduction to Social Data Science</span>
]
.author[
### David Garcia<br><span class="semester">Winter Semester 2026/27</span>
]

---

layout: true

<div class="unikn-footer"><span class="fbar-left">Session 1</span><span class="fbar-mid">Introduction to Social Data Science</span><span class="fbar-right">Winter Semester 2026/27</span></div>

---

# Today

1. Introduction to Social Data Science
1. The Parable of Google Flu Trends
1. Measuring Temporal Orientation with Google Trends
1. Course logistics

---
class: section, center, middle

# Introduction to Social Data Science

---

# What is the aim of Social Data Science?

> The aim of Social Data Science is   
**the Quantitative Understanding of Human Behavior**.

- **Quantitative:** As opposed to qualitative or descriptive, we aim for robust findings grounded in strong evidence that can be quantified.

- **Understanding:** Not just predicting, we want to be able to generalize and combine knowledge, and even to motivate interventions or policies.

- **Human:**  We will not study particles or objects. Measurement validity and ethics will be a challenge.

- **Behavior:** Observable changes, structures, dynamics, and patterns; not just stories or theories

---

#  How are we going to do it?
**Retrieving, processing, analyzing, and interpreting Digital traces.**  
<center>
![:scale 80%](FBMap.png)
</center>
---

# Another example of digital traces
.pull-left[
**Pair up and guess together:**

- What kind of traces are shown in the figure on the right?

- How were they collected?

- What can we learn about human behavior with this kind of data?

]
.pull-right[.center[![:scale 100%](Net.png)]]

---

# What is data?

> **Data:** Facts in the form of stored and transmittable information.

- *Data* is the plural of the Latin word *datum*. *Data* means “given (things)”.  
- Data is given to us, it is neither fabricated nor simulated.
![:scale 85%](http://www.phdcomics.com/comics/archive/phd080715s.gif)

---

# What is Social Data Science?

**Data Science** is the application of methods from Computer Science and Statistics to empirical questions and practical problems.

**Social Data Science** is the application of Data Science to study human behavior and social interaction.

- Importance of combining methods, synthesis, and communication
- Gathering digital traces with Computer Science techniques
- Interpretation of analyses with respect to what is known in the relevant Social Sciences
---

background-image: url(VennV2-1.svg)
background-size: 97%
---
background-image: url(VennV2.svg)
background-size: 97%
---

# Why digital traces?

Data from digital traces have six properties that can complement other traditional data sources in the social sciences:

- **Big Data:** Observing large amounts of humans across demographics

- **Fast Data:** Quantifying aspects of human behavior in real time

- **Long Data:** Retrieving longitudinal data and at various timescales

- **Deep Data:** Gathering persistent information on individuals

- **Mixed Data:** Combining heterogeneous data sources and unstructured data

- **Strange Data:** Locating small subcommunities or deviant behavior

---

# What are the limitations of digital traces?

With the great potential of digital traces also come great challenges:

- **Platform biases:** caused by their design, algorithms, social bots, etc

- **Data gatekeepers:** Not everyone can access some data sources

- **Performative behavior:** Talking online is not the same as offline.

- **Representativity issues:** Not everyone leaves digital traces.

- **Observational data:** Testing causal mechanisms is not straightforward.

- **The data deluge:** Too much data enables black-box predictions that can be useful but limit our understanding.

---

## The Hype Cycle of Social Data Science
<center>
<img src="Hype1.svg" alt="" width="900" />

---

## The Hype Cycle of Social Data Science
<center>
<img src="Hype3.svg" alt="" width="900" />

---

### The importance of questions in Social Data Science

![:scale 100%](DeepThought.jpeg)  
Deep Thought, from the movie version of The Hitchhiker's Guide to the Galaxy.

> **Understanding our questions is a prerequisite to understanding their answers.**

---
class: section, center, middle

# The Parable of Google Flu Trends

### A Social Data Science Story

---

# Google Trends

Google Trends is a website hosted by Google that allows you to get a measurement of Google search volume for a term

<center>
<img src="EU.png" alt="" width="900"/>  
</center>
---

## Nowcasting flu incidence with Google Trends

**Nowcasting** is predicting the present. It provides an estimation of the value of a quantity based on signals that appear at the same time.

Google Flu Trends aimed at nowcasting influenza-related physician visits based on Google search volumes.

As reported in the [Nature paper](http://www.nature.com/nature/journal/v457/n7232/full/nature07634.html), Google Flu Trends achieved a high weekly accuracy between 2003 and 2008.

The figure shows the nowcasting result of Google flu trends. CDC data is published with a delay of two weeks.

---

## When Google Flu Trends Stopped working

---

## When Google Flu Trends Stopped working

In 2013, Google Flu Trends [started overestimating](https://www.nature.com/news/when-google-got-flu-wrong-1.12413) the incidence of flu cases.

A [follow-up study in Science](https://doi.org/10.1126/science.1248506) showed that just a lagged CDC prediction with 2 week-old data outperformed Google Flu Trends, without using any Google data.

Since many details about how Google and Google Trends work are not public, we cannot know for sure what the source of the error was. Some candidates are news-related searches, seasonal effects, or changes in Google's algorithm and interface.

Google Flu Trends is an example of **Big Data Hubris**: “The often implicit assumption that big data are a substitute for, rather than a supplement to, traditional data collection and analysis” [(Lazer et al, Science, 2014)](https://doi.org/10.1126/science.1248506)

> Take home message: **All data is better than Big Data**

---

## Beware of Making a Data Piñata

![Definition of data piñata from the [Urban Dictionary](https://www.urbandictionary.com/define.php?term=data%20pi%C3%B1ata)](Pinata.png)

Also known as kitchen-sink regression, garbage-in garbage-out, and post hoc storytelling. Some patterns will always come out with sufficient data, what matters is making sense of them.

---
class: section, center, middle

# Measuring Temporal Orientation with Google Trends

---

# Time, culture, and the economy

What is the relationship between culture and the economy?

![:scale 150%](Culture-Economy.png)

---

## Long-term orientation and economic development
> **Long-Term Orientation (Geert Hofstede)**  
Long-term oriented societies believe that the most important events in life will occur in the future; short-term oriented societies believe that those events occurred in the past or take place now.

.center[![:scale 60%](FOIGDP.png)]

---

## Measuring the Future Orientation Index

[Preis et al. 2012](https://www.nature.com/articles/srep00350) proposed a way to measure how much a society looks towards the future with Google Trends, the Future Orientation Index (FOI). The FOI for a country *c* in year *y* is calculated as:

.center[
`\(FOI_{c,y} = \frac{G(y+1,y,c)}{G(y-1,y,c)}\)`
]

where `\(G(y_1,y_2,c)\)` is the Google Trends volume for searches for year `\(y_1\)` during year `\(y_2\)` from country `\(c\)`.

The FOI measures the ratio of search volume from a country for next year divided by the search volume for the previous year in the same country.

---

# Example of trends for FOI

.center[![:scale 85%](FOI1.png)]

---

# Example of trends for FOI
.center[![:scale 85%](FOI2.png)]

---

# Exercise 1: Do richer countries look more to the future?

In [Exercise 1](../../Session03_RetrievingData/Exercise1/GDP_FOI_Short.html) (from Session 3), you will calculate the Future Orientation Index with Google Trends and test if it is still correlated with GDP per capita.

.source[[Preis et al. (2012)](https://doi.org/10.1038/srep00350), *Scientific Reports* 2:350. Images: CC BY-NC-SA 3.0]

.pull-left[
.center[![:scale 90%](Preis2012_page1.png)]
]
.pull-right[
.center[![:scale 62%](Preis2012_Fig1b.png)]
]

---
class: section, center, middle

# Course logistics

---

# How this course works

- **13 weekly sessions** of 90 minutes, Mondays from 19.10.2026 to 1.02.2027 (no class on 30.11.2026)
- **Lectures with slides** on Social Data Science topics, and **practical sessions** where we work together through R Markdown notebooks
- **Exercises** to apply what you learned, about one every two weeks
- **Final project** of your own: the basis of your grade

All materials (slides, notebooks, exercises, data) are on the course page: <https://dgarcia-eu.github.io/SDS_Intro_Konstanz>. They may be updated during the course, so download them again before each session.

> **Homework for Session 2:** install R and RStudio (see the last slide of today)

---

# Exercises

Each exercise starts at the end of a practical session. You continue it at home, and we discuss the results together in a later session.

| Exercise | Topic | Starts | Discussed in class |
|---|---|---|---|
| 0 | BMI and daily steps (practice) | Session 2, 26.10 | |
| 1 | Future orientation and GDP | Session 3, 02.11 | Session 6, 23.11 |
| 2 | Division of social impact on Bluesky | Session 6, 23.11 | Session 9, 21.12 |
| 3 | Evaluating sentiment analysis | Session 8, 14.12 | Session 10, 11.01 |
| 4 | The retweet network of Swiss politicians | Session 10, 11.01 | Session 12, 25.01 |

The exercises are **not graded**, but they train exactly the skills you need for your project.

---

# Submitting exercises for bonus points

You can submit your solutions to Exercises 1–4 to improve your grade:

- **2 or 3 solutions that pass:** bonus of **0.3** on your grade
- **All 4 solutions pass:** bonus of **0.6** on your grade

**Deadline:** the day before the exercise is discussed in class (22.11, 20.12, 10.01 and 24.01).

**How:** send the R Markdown file of your solution and its knitted HTML by email to david.garcia@uni-konstanz.de. Make sure that it knits without errors.

---

# Grading: the final project

Your grade is based on an individual project with two parts:

1. **Report** of max. 6 pages (min. font size 11pt), with the R code that generates the results attached. It gives the **numeric grade** of the course (plus the exercise bonus).
  - **Deadline: 15.03.2027**, submision over email to david.garcia@uni-konstanz.de. 
  - Please only pdf, not word or anything else.
2. **Online presentation with Q&A** about your project: pass/fail (Studienleistung), necessary to pass the course. It takes place after the report submission, **between 23.03 and 31.03.2027**.

You can get input on your project ideas and plans in the lectures and in consultations you book by email, until two weeks before the report deadline. **Ask for input early!**

Optionally, send a **short project description** (motivation and project plan) by email to get structured feedback.

---

# Structure of a project report

A good Social Data Science project has seven parts:

1. **Motivation:** What question do you seek to answer and why?
2. Project plan: Which tasks are needed to answer the question?
3. Data retrieval: Collect the data with interfaces and other resources
4. Data processing: Filter data, normalize values, compute variables
5. **Analysis:** Statistical analyses and visualizations that assess the question
6. **Conclusion:** Answers to the question and how reliable they are
7. **Critique:** Limitations and alternative explanations

The parts in bold weigh more in the grade, but all seven have to be reported. Extra points for open science (e.g. code and data on GitHub) and for data or methods beyond the course.

---

# Start with a question

- Social Data Science is **not data-driven**: start with a question about human behavior, not with a cool dataset
- Good questions have a **yes/no answer or a number**: "Are emotional posts reposted more often?", "Is the Future Orientation Index stable across years?"
- Questions starting with "how" or "what" are usually too vague: "How do people use TikTok?"
- Commit to your question early, and document changes of data or methods in your report
- Results that do not fit your expectations are fine: what matters is that you apply the right methods to the right data
- Projects are **individual**. Talk with your colleagues, but do not copy code from online resources or coding bots: you have to understand and explain your code in the presentation

---

# Examples of past projects

.pull-left[
**Popularity assortativity of musicians on Spotify**
- Network of musicians in the same songs
- Assortativity of popularity, with a permutation test

**Social Impact Theory and COVID-19**
- Impact: Google Trends and tweets
- Immediacy: distance to early outbreaks
]
.pull-right[
**Sentiment about Trump and regional income**
- Tweet sentiment per region with Syuzhet and VADER, validated by manual annotation
- Regression with census income and education

More examples and answers to frequent questions in the [projects guide](https://dgarcia-eu.github.io/SDS_Intro_Konstanz/ProjectsGuide.html)
]

---

# Homework before Session 2: install R and RStudio

Session 2 is an R crash course that starts directly with R, so please prepare your computer before it. It takes about 30 minutes. Follow the [setup instructions](../../setup/Setup.html):

1. Install **R** from <https://cloud.r-project.org/> (version 4.1 or newer)
2. Install **RStudio Desktop** from <https://posit.co/download/rstudio-desktop/>
3. Download the course materials and open `SDS_Intro_Konstanz.Rproj` in RStudio
4. Install the course packages: `source("setup/install_packages.R")`
5. Knit `setup/SetupCheck.Rmd`: it should end with "Everything is ready"

> **If something does not work:** use the **Posit Cloud** fallback described in the setup instructions. It runs all course materials in your browser, so you can follow Session 2 anyway.