Problematic 3D Representations

Molyneaux declares "The reinforcement of ideas in some image is very powerful... Picture and other visual representations...have a tremendous inertia, or staying power, that might persist long after the ideas behind the images have gone out of fashion" [1]. And this is likely true to an even greater extent of 3D representations.  This suggests that creators of 3D representations ought to consider with great care their choices in designing and executing such representations, for potential negative effects of missteps are amplified when one is working with 3D representations.  Here we focus on the potential for missteps in creating 3D virtual representations of environments and also in creating the occupants of such environments (avatars and indigenous occupants of such environments ("bots").  Likely, the points we make below can be extended to other sorts of 3D representations as well (for instance, 3D representations of artifacts, or of buildings), but we do not develop this extension here.

3D representations of environments:  lessons from a trolley problem experiment

Philosophers, social scientists, and communications scholars often generate data by posing what are called "thought experiments" to their audience (a scenario which a subject is asked to imagine and respond to). These experiments are almost always presented linguistically, through text or spoken language. This allows the presenter of the thought experiment to leave many details of the scenario unrepresented. However, might these unrepresented details matter? It appears as though the unrepresented details of thought experiments may affect the intuitions of the recipient of the thought experiment.

In the traditional presentation of the trolley problem, you are asked to imagine yourself standing at the pictured switch, and to consider whether you would let the trolley continue on its track to run over the five individuals on the left track, or divert the trolley to run over one individual by pulling the switch; sketch of the trolley problem.  Preliminary results from an implementation of the trolley problem in a 3D virtual environment suggest that how one concretizes or fixes previously unrepresented details can influence the judgments and decisions of the participant.

One consequence or takeaway for DH researchers in general, and for DH researchers working in 3D or 2D in particular, is that how one fills in the details of a representation may have unintended effects. This is a special challenge for DH because 3D and 2D environments typically force a researcher to "fill in details" (when representing a scenario with language naturally allows for ambiguity).

3D representations of avatars and autonomous computational occupants of virtual environments (bots):  lessons from a sampling of past efforts

If constructing 3D virtual environments offers opportunities for 3D content creators to generate unanticipated and potentially negative effects, then surely populating such environments offers even more such opportunities.  Populating 3D virtual environments is important in general because most of the spaces we are interested in representing contain people, which implies that adequately representing spaces frequently requires adequately representing people.  When creators of 3D virtual environments seek immersion as a goal, choices about how to construct avatars are particularly important, as the experience of moving around a space (physical or virtual) is intimately connected to having a body (physical or virtual!).  But the dangers of populating 3D virtual environments with avatars and/or computational autonomous agents (a.k.a., bots) are legion, ranging from relatively minor dangers like jarring explorers out of immersion in an 3D virtual environment to more significant dangers like deeply offending people or buttressing past social inequalities and injustices.  

Here's an example of an avatar choice selection likely to jar people out of becoming immersed in the 3D reconstruction of Arkansas State's historical Lakeport Plantation property (Lakeport Plantation shown in the background; image courtesy of Alyson Gill).

The avatars at the Lakeport Plantation slide was created to make a point, but we can find jarring instances of avatars occupying actual 3D immersive environments.  For a time, Edward Gonzalez-Tennant maintained a virtual museum of the town of Rosewood, Florida in the MUVE Second Life.  The mostly African-American town of Rosewood, Florida, was destroyed by a mob of angry whites in early January, 1923.  Six African-American residents of Rosewood were brutally killed during what is now called the Rosewood Massacre and the and virtually the entire town put to the torch.  Follow the link provided and scroll down the page to view a still image of a crowd of avatars visiting Gonzalez-Tennant's Second Life Virtual Rosewood Museum.

Another instance of an avatar that is potentially jarring in a different way occurs in Emily Roxworthy's 2013 interactive 3D model of a Japanese-American internment camp once located in Jerome, Arkansas, Drama in the Delta.  Players in the game assume the role (and guise) of Jane, a 14 year-old Japanese-American girl confined a the camp.  The Jane avatar underwent a number of revisions, in response to various criticisms.  Please follow this link to view a later iteration of the Jane avatar (and see if you can tell what some of the fuss might have been about.)

There are digital humanities projects that have taken considerable care in the design and deployment of avatars.  One example is the Digital Hadrian's Villa Project, directed by Bernie Frischer of the Virtual World Heritage Laboratory at Indiana University.  In the 3D interactive Unity version of the villa, those wishing to explore the villa are presented with a curated set of avatar options representing the various sorts of people one might have found in the villa in its heyday:  slave, courtier, freeman, scholar, soldier, senator, even emperor.  The different avatars are dressed differently (according to the best scholarly reconstructions), but more interestingly, they have differential access to different areas of the villa (corresponding to rank).  By thinking about this system, we can perhaps get a sense for how avatar design and deployment might be a force for good, not just by conveying historically accurate representations but also by giving those animating these avatars a first-person sense for what it might be like to inhabit various social roles with which they might have little experience, that of slave, for instance.  Above is a screenshot of the avatar choices in the Unity version of the Digital Hadrian's Villa Project, image uploaded from idialab.org.

Group activity

Pair up with another member of our institute.  Choose one of your DH projects.  Think about the implementation of the project you have chosen, and discuss the most prominent moral challenges you see as being involved in this implementation.

Alternatively (if neither of your projects is well-suited for this), view an episode of the web series Ask a Slave (your choice of which episode).  Imagine you are charged with translating this episode into a 3D virtual interactive environment, complete with avatars (and possibly autonomous computational agents, a.k.a., bots).  Discuss the most prominent moral challenges you see as being involved in doing so.

This page has paths:

This page references: