1 of 37

Module R561

Scientific Methods

Hochschule München University of Applied Sciences

Marloes Geboers, PhD candidate�Digital Methods Initiative, University of Amsterdam

Visual Methodologies, AUAS

Department 14

2 of 37

R561 Scientific Methods

R561 - Scientific Methods

After the lecture you…

understand the foundations of digital methods and what role such methods can play in tourism management

are familiar with the methodological and practical issues when ‘doing digital methods’

can reflect on the advantages and disadvantages of digital methods

2

3 of 37

R561 Scientific Methods

R561 - Scientific Methods

24

Disciplines in tourism

Disciplinary thought is often linked to worldviews, philosophies, ideologies, streams of thought

….AND METHODS!

Philosophical worldviews

Research designs

Research methods

Research approaches

Quantitative Qualitative Mixed Methods

Postpositivist Constructivist Transformative Pragmatic Critical realist

…

Questions Data Collection Data Analysis Interpretation Validation

Source: Cresswell, 2014, p. 35

Digital methods

4 of 37

What do digital methods researchers do?

‘We follow the medium’

5 of 37

Digital Methods and Tools for Social Research

@ Digital Methods Initiative, UvA

www.digitalmethods.net

6 of 37

Gamification of sociality

On sites like Twitter, everything anyone says is gamified with scorekeeping through quantified retweets, faves, and followers. To make so much of our sociality permanent, categorized, and explicitly and quantifiably ranked into hierarchies produces not sociality for its own sake but one that is concerned with success and failure according to [..., social] platforms.

Nathan Jurgenson, The Social Photo (2020)

7 of 37

when in tourism management….

Question: what can you do with all those retweets, faves, likes and followers?

8 of 37

HOW IS THE VIA FRANCIGENA TALKED ABOUT ON INSTAGRAM?

CITY MARKETING, PERCEPTION OF CITIES ONLINE

9 of 37

INSERT EXAMPLE MüNCHEN TOURISM

10 of 37

Quali-quanti nature of digital methods

11 of 37

outputs...

12 of 37

hashtags as powerful demarcators of data

13 of 37

buttons as proxies for emotions?

14 of 37

again… very quali-quanti

15 of 37

16 of 37

17 of 37

Digital objects and what they ‘let us do with content’

content

digital ‘objects’ allow us to do things with content: share, like, follow, tag etc, but they are also limiting us (as researchers and as users!)

18 of 37

Limited access

19 of 37

But Marloes: how can we then get at this fabulous data?

web scraping (open to semi-closed to closed)

for social media we have to tap into APIs

20 of 37

Wait, what?

21 of 37

methodologically, you may regard the API as ‘the boss’

22 of 37

Methodological considerations when engaging in following the medium

Twitter images of the Syrian war organized by�Google Vision API

I sized the images according to retweets

Class question:

What is assumed when using metrics such as retweets?

23 of 37

Risks when interrogating APIs..

https://smart.inovamedialab.org/smart-2019/project-reports/interrogating-vision-apis/

Commercial services like Google Cloud Vision API, Microsoft Azure Computer Vision API and IBM Watson Visual Recognition API are considered “black boxes” (Pasquale, 2015; Buolamwini & Gebru, 2018), and the auditing of their precision or coverage is hampered by the lack of transparency..

Regarding the computer vision providers, the study shed light on their differences, limitations and ways of repurposing them for exploring culture and representations embedded in digital platforms with digital methods. Commercial services like Google Cloud Vision API, Microsoft Azure Computer Vision API and IBM Watson Visual Recognition API are considered “black boxes” (Pasquale, 2015; Buolamwini & Gebru, 2018), and the auditing of their precision or coverage is hampered by the lack of transparency. Neither the list of labels, tags or classes nor their total amount are disclosed by the providers. We hope studies like ours add up in the pool of cases to help other researchers interested in using computer vision APIs to social research. Also, different API providers don’t overlap or match their understanding of complex cultural data such as images, so using them uncritically can be a methodological problem.

The tentative categorization of the image tagging coverage and quality (“rookie”, “amateur”, “expert”) between the three providers does not mean that one of them should be preferred nor that researchers should use only one in their projects. Some of the procedures performed during our study (e.g. clustering dictionary) shows the feasibility of combining two or more providers merging the annotated datasets and networks. Further experimentation with other AI providers is recommended to broaden the scope of the discoveries.

24 of 37

SMART Data Sprint: Beyond Visible Engagement

28 January – 1 February 2019

Universidade Nova de Lisboa | NOVA FCSH | iNOVA Media Lab

25 of 37

The context of war..

earthquakes

26 of 37

tourism, recreation and advertising

27 of 37

tourism, fun, girl, street

28 of 37

29 of 37

Images too small to see….

sleep, nap, arm

chest, beard, zombie

30 of 37

Important considerations

Online content has a certain ‘technicity’;
Online content is co-produced by platforms and engines;
Platforms may offer windows to the world of a cultural or social issue;
Not all online data is public...

A fifth point worth mentioning is that while Wikipedia offers public views on its technicity, the other platforms studied in this book do not. Google Web Search, through its terms of service, does not allow for the use of its search engine for anything other than search. So repurposing the

engine as a research device (as discussed in detail by Weltevrede 2016) goes against its rules and regulations. Twitter has various APIs, however, on an interface level, Twitter disclose its mechanisms of ranking and prioritizing content (and neither does Google or any social media platform). Current FB scandals underline the importance of developing a thorough understanding of the platforms and ethical methods to study their content, building on strong traditions from the humanities.

The meth consequences of working with APIs (what are they hihi) see paper Bernhard on FB API

Niet te diep erin

Black box problems, algorithms: is popularity also visibility? We simply cannot know

Example: showcase my work with retweets, most liked etc.

Instability of research objects (removal by users, deplatforming, banning of accounts etc) but also instability due to a constant recontextualization of content (different attribution of meanings)

THE WORK of https://medium.com/1st-draft/searching-for-the-misinformation-twilight-zone-63aea9b61cce distortion because of like/follower ratios (merely looking at the top is missing a huge part of the story, skewed reality due to popularut metrics, refer shortly to Jurgenson and social media (scorekeeping and gamifying content)

31 of 37

co-production of platforms that standardize sociality

What is the API? The data boss….

32 of 37

Advantages

�

Features of platforms can be creatively repurposed for research

You can get to know your region or city or issue in a different way

Wealth of knowledge produced by a huge number of people that tweet, tag, post and like

33 of 37

Disadvantages

Very API dependent / You can’t always get what you want

Are you researching the social issue or how the platform standardizes it? / blackboxes of algorithms

Limited knowledge of motivations and incentives of people to like, share etc.

Instable research objects (users recontextualize content constantly and platforms ban or censor content or even break data)

34 of 37

what can digital methods bring to tourism management research?

Digital methods allow for new ways of visualization of sociality playing out on social media

and new ways of getting to know your ‘subject matter’

E.g. in city marketing: how do people perceive a city? Or the problems that tourists pose to cities?

You may learn more about your own impact on social media (campaigns)

Impact analysis of campaigns

Perceptions of cities (placemaking is already done by people amateurs on socials)

35 of 37

> pointer to the tools that we use & mutual learning exercise

What kind of data would you want to get from these tools and how could it help your field?

https://wiki.digitalmethods.net/Dmi/ToolDatabase

CrowdTangle �(precurated by Facebook! though still millions of accounts and pages are archived)

36 of 37

Further literature and inquiries:

=

26

Digital Methods (Rogers, 2013)

Doing digital methods (Rogers, 2019)

Why buttons matter (Geboers, Scuttari, Stolero & Teernstra, 2020)

37 of 37