Technological Tact in Tactical Iraqi
Published by Mark Marino November 23rd, 2005 in Uncategorized, HCTI, Poetics, bots, Text Art, Criticism, HAI.Interactor: How do you ask an Iraqi where the warlord is so you can meet with him to chat?
Iraqibot: Very politely.
“Tactical Iraqi,” W. Lewis Johnson’s language instruction program, offers soldiers the opportunity to learn Iraqi through interaction with characters through a video game-like interface. Johnson is current director of the Center for Advanced Research in Technology for Education (CARTE) at the Information Sciences Institute of the University of Southern California.
According to Johnson, the response from Iraq War veterans has been outstanding to the effect of “I wish I had used that program BEFORE I went over there.” Their site reports, “Learners at Grafenwoehr who had previously been deployed to Iraq stated that they learned more Arabic in one day working with Tactical Iraqi than they learned in their entire tours of duty in Iraq.”
The system, as a teaching tool, is impressive. Dr. Johnson was polite (and generous) enough to share it with me on a recent visit to his lab in Marina Del Rey. The voice recognition was strong and the interaction compelling. (Nonetheless, I don???t think it will count as my military service.)
But the system isn’t merely utilitarian, Johnson stresses. It does more than teach pronunciation (apparently I mastered several Iraqi greetings within minutes!). The system stresses decorum, which Johnson ties to cultural literacy, human decency, effective chatbot design, and, well, tact. Some of the reporting on Tactical Iraqi has taken up this rhetoric of “Cultural Interaction, Not Fighting” and after a traumatic trip to E3 this summer, I felt sympathetic to the call for civility, especially in war-related gaming.
As he explains, what’s bothersome about some autonomous help agents is that they intrude, they bother, they pester. This is no way to help. Certainly students do not respond well to this kind of pedagogy by humans, let alone bots. (See his paper* for more on this).
But what does it mean to be polite?
In “Tactical Iraqi” if the interactor does not use appropriate gestures and greetings, the other characters will not trust them (as indicated by a meter above the person’s head). Seeing this systematization of cultural sensitivity in action, I was impressed by the desire to build a chatbot system that is predicated on civil exchange, while so much interaction is based on abuse. When you arrive at the house of the man you seek, you must first remove your shoes. Find me the first person shooter, built by the Army or otherwise, that has such respect. And yet the respect can also be troubling…
Of course, “Tactical Iraqi,” or rather the Tactical Training Language System project “is sponsored by the DARWARS Training Superiority Program of the Defense Advanced Research Projects Agency in cooperation with the U.S. Special Operations Command, the U.S. Military Academy, the U.S. Marine Corps, and other government agencies. ”
Again what does it mean to be polite if you are a soldier in a simulation that is training you to get other people to yield to you information? What does it mean when they are not other people but models of other people designed in the country which maintains a controlling military presence? What does it mean to be compelled to be polite by a system that trains you how to compel others to submit information?
When you play the game, you become an agent with a State data command and control operation. you are collecting data following from agents using appropriate (unobtrusive) protocol, marshaling data by interrogating people using appropriate techniques.
Is this teaching how to be polite and culturally aware through a language-training system or developing adequate models of the subject peoples and their speech in order to control them or to train others to control them through locating noncompliant members of the community?
I do not mean to be impolite with this question or to downplay my respect for this project. This is not the “tactical” of the Advanced Tactical Fighter. This is all a far cry from shooting terrorists with sniper sights, but it is a testament to the potential insinuations of produced consent that a civil game might hide beneath the long-flowing robes of “custom.”
All of these issues or more complicated, I know. So I pose a few questions:
What are some other thoughts on this? May I politely ask for other writings on the matter?
How does the creation of autonomous agents relate to the work of imperialism?
Where else does Imperialism come up in the world of conversational agents?
*”Experimental evaluation of polite interaction tactics for pedagogical agents.” Ning Wang, W. Lewis Johnson, Paola Rizzo, Erin Shaw, Richard E. Mayer
January 2005 Proceedings of the 10th international conference on Intelligent user interfaces
I also had a feeling of uncomfortable complicity when approaching “Tactical Iraqi” as an object of study. Like Mark, I found the research team accessible, intelligent, and thoughtfully responsive to tough questions. Nonetheless, I listed about a dozen serious reservations about the game in a recent paper for the Digital Arts and Culture Conference. Now I have found one more reason to be skeptical. What struck me in yesterday’s Los Angeles Times opinion page was Max Boot’s article, “Navigating the ‘human terrain,’” in which Boot, a senior fellow at the Council on Foreign Relations, enthuses about visiting “the Expeditionary Warfare School, where captains study Arabic by playing a sophisticated computer game complete with animated characters.” Boot argues that this provides critical training that simulates “the human terrain” of the theater of conflict in Iraq. What I find troubling after reading Boot is coming to terms with the fact that Tactical Iraqi may be winning so much media praise (notably from NPR and the New York Times) and earning a top DARPA award not because it is an effective way actually to teach Arabic but because it provides an easy interface to SHOW the teaching of Arabic to U.S. soldiers to the public as a form of display. It enables a kind of spectacle of cultural sensitivity aimed at outside audiences that traditional classroom learning can’t provide. Thus the learner can become a virtuoso performer in Baudrillard’s Simulacrum.
Elizabeth,
I look forward to reading your paper. So do you see this show as a theater of US cultural production to win the hearts and minds of US skeptics? That’s interesting.
Part of my trouble is that I’m torn. Having seen the system and having learned a few words through it, I must say that I like it as a teaching system. I like that it isn’t a shooter. Also, I like the words “cultural sensitivity.”
The trouble is, cultural sensitivity becomes a new kind of operational system in the command and control schema of larger military objectives.
This quote from the presentation Boot quotes suggests the way cultural knowledge can be operationalized.
Perhaps a pragmatist would be more approving: i.e. “Your chatbots can’t HANDLE the truth.”
When I gave my DAC 2005 paper on Tactical Iraqi, many in the audience said that they liked the paper but took issue with my conclusion, which read as follows:
“As a rhetorician who studies new media, I sincerely hope that this military simulation, which is intended to forestall violence and armed conflict rather than prepare for it, will not be the last game in its genre. Although I have expressed concerns about its technological and ideological hard wiring, I think that this program also stimulates critical discussion about how trust and self are constituted by digital experiences and how language is integrated into virtual environments of all kinds, even when conventions about rules and randomness in virtual spaces do not match up neatly with the principles governing linguistic play.”
I was surprised to be characterized as an apologist for saying that the game represented a worthwhile genre, particularly when my reading was so critical on so many fronts. But it’s interesting that you mention pragmatism, because I suppose my thinking about Tactical Iraqi has been really pragmatic up to this point, which leaves me open to accusations of philosophical amorality.
On the one hand, so many Iraqis — 30,000 according to our Chief Executive — have died in this war, and many of these deaths are painfully ironic because they were caused by military personnel shouting at civilians in incomprehensible English rather than passable Arabic.
Yet if you argue from a utilitarian perspective that it’s most important to reduce the number of civilian casualties that result from military occupation, you can find yourself on very shaky ethical ground. For example, I certainly wouldn’t argue that if the U.S. Military had a system that could dispense sleeping gas from robotically controlled planes, it would be a more moral military instrument. Even if such a system would cause geopolitical coercion to be considerably less lethal, people can also be harmed when their capacity to resist is stymied.
Lately, I’m also going back a kind of stasis theory reading of Tactical Iraqi. The project is justified by two problems facing the army: 1) a chronic shortage of Arabic speakers and 2) a confusing theater of conflict that overlaps with civilian geography. It seems like these problems could be more efficiently solved through organizational and political means.
As a rhetorician who believes in the importance of individual speech acts and collective communicative action, increasing the number of potential translators in the world seems like a legitimate goal. I’m just not sure if Tactical Iraqi is the way to do it.
Sorry I’m commenting so late on this thread, but I saw Elizabeth’s excellent DAC presentation, and I think it captured the problem of spectacle quite well. To generalize to other forms of programmed language interaction and bot fiction -
Are there general advantages of a contextual spectacle - that is, if Berlitz did a Learn Spanish game with levels like “At the restaurant” and “in the airport,” would it help knowledge acquisition?
In terms of fiction, are conversations with spatially located and contextualized bots generally better in part because they put automatic limits on what topics or modes are situationally appropriate? One of the hobgoblins of chatbot design is the open-ended nature of potential human discourse - but “tact” requires context, which bots are generally quite bad at delivering….