Conversation Visualization

Tracing interdisciplinary design conversations

Collaborator - Katherine Giesa, Emek Erdolu
Completion - Fall 2020


D3.js
Natural Language Toolkit
python



With the increasing convergence of digital and architectural design practices, architects and tech designers engage in interdisciplinary conversations more often.

However, how effectively do architects and tech designers communicate? How do they engage through the discourses in a conversation? Through text-based data visualization, we hoped to extract more insights about the engagement.
Context

During the 2014 Venice Biennale, German architect and journalist Niklas Maak, renowned Dutch architect Rem Koolhaas, and American engineer, tech designer and entrepreneur Tony Fadell, the founder of Nest Labs, had a heated exchange of perspectives on the digitalization of architectural elements and the Nest's self-learning thermostat.


Analysis
Conversation
Overview︎︎︎
As the actor-time heatmap shown below, Fadell had continuous participation in the conversation while the architects Maak and Koolhaas took turns challenging Fadell about his points.

Terminology︎︎︎

Function words maintain the syntactic structure in a sentence but do not contain semantic information, such as prepositions, pronouns or conjunctions.
Content words are words that possess semantic content and contribute to the meaning of the sentence in which they occur.


Weight of
Conversation︎︎︎
Lexical density is defined as the number of content words divided by the total number of words, which is a measure of how informative a text is.
While Fadell has the most weight in the conversation in terms of the total number of words, Maak has the highest lexical density in his discourse, which correlates to Maak's role as the moderator in the conversation.


Language Style
Matching︎︎︎
The three network graphs show function words as nodes and its frequency by each actor as links of different thicknesses. Despite having less and shorter dialogues through the conversation, Maak and Fadell had a slightly higher degree of engagement through their discourses.
Language Style Matching (LSM) uses the degree of match between the number of function words in actors' discourses as a factor of engagement in a conversation. (Gonzales et. al, 2010) From the bar graphs of LSM, we see similar dyadic patterns between Koolhaas and Fadell, and between Koolhaas and Maak, which suggests there are more engagement Koolhaas-Fadell, and Koolhaas-Maak.
Disciplinary Word
Glossary︎︎︎
The word tree map displays catergories of disciplinary wordds in different colors. The sizes of the colored clusters indicate there are more inter / dual-disciplinary words used than the disciplinary words.
Conversation
Exchange︎︎︎
The actor-word network graph reveals the shared content words and the frequency of the words. Koolhaas and Maak have a more balanced mixture of content words, while Fadell heavily leans towards technological design words and interdisciplinary words. Overall, technological design words are more commonly used than other types of content words.

This pattern matches with the technological-design-focused dynamic where Koolhaas challenges technological design practices over privacy from an architectural perspective. At the same time, Fadell actively defends his position and rarely gets to engage with architectural topics.
︎meijieh@andrew.cmu.edu