of the American Society for Information Science and Technology     Vol. 28, No. 3      February / March 2002

Search

Go to
Bulletin Index

bookstore2Go to the ASIST Bookstore

 

Copies

Annual Meeting Coverage

ASIST 2001 Plenary Debate

James Hendler and Ben Shneiderman on the Next Generation of Interfaces

by Steve Hardin

James Hendler is former chief scientist of the Information Systems Office at the Defense Advanced Research Projects Agency and professor at the University of Maryland, College Park. He can be reached at hendler@cs.umd.edu.. Ben Shneiderman is professor in the Department of Computer Science at the University of Maryland, College Park. He can be reached at ben@cs.umd.edu .. Steve Hardin can be reached at libhard@cmi.indstate.edu .

What should the future Web look like? How do we get there? To what extent should the Web mimic human behavior? Two experts squared off on these and other issues before several hundred audience members during the closing session of the ASIST Annual Meeting in Washington on November 8, 2001.

Organizers billed the session as a debate between Dr. Ben Shneiderman, a computer science professor at the University of Maryland, College Park, and also the founding director of the Human-Computer Interaction Laboratory; and Dr. James Hendler, former chief scientist of the Information Systems Office at the Defense Advanced Research Projects Agency (DARPA), as well as a professor at the University of Maryland, College Park. Both men presented their ideas on "The Next Generation of Interfaces: Responsibility and Control," and discussed their differences with each other and with the audience. ASIST President-elect Trudi Bellardo Hahn moderated the session.

Shneiderman began his presentation by saying that we'll build cognitively comprehensible systems, giving us a sense of mastery, control and satisfaction. In the process, we'll take control for the results of our decision. He envisions systems that are "affectively acceptable." But he hastened to add they would not be adaptive, autonomous and anthropomorphic. He said anthropomorphic designs persistently fail.

The scientific approach, he said, extends beyond "user friendly" systems. We want to specify users and tasks, predict and measure, he said, and focus on the future of Web use. There are three directions we need to consider:

  •   Effective visual displays. Approaches like Spotfire and Treemap will empower us.
  •   Social dimensions. They're essential to support online communities. Human-human communication is central. Trust is key. We can invite participation by ensuring trust. "You can rely on machines," Shneiderman said, "but you trust people. When you blur the boundaries between people and machines, you run into trouble." Machines can't really be responsible for anything, he said.
  •   Universal design. We need to support a broad range of users, he said, including bridging the gap in users' knowledge. Online help isn't enough.

Shneiderman wrapped up by listing some of his disagreements with Hendler:

  • Importance of agents vs. human interaction. Shneiderman thinks the notion of "intelligent agents" misleads people into thinking the machine will do the job for us. The idea of "intelligent" even "spiritual" machines has not helped. But he sees hope in support of the notion of human dialog: "Tell me what you have in mind."
  • Shared vs. multiple ontologies. He noted there's a difference in the meaning of "the secretary" in a typical office and the same phrase in the Pentagon. We need to find a middle ground supported by dialog.
  • Empirical testing vs. "build it and they will come." Shneiderman said we must find out what people need and want before we build. The delay in consideration of human issues leads to the construction of the wrong structures.

Shneiderman finished his presentation with a quote from Thomas Jefferson about his desire to see knowledge so widely disseminated that it reaches beggars as well as kings.

Hendler began his part of the session by noting Shneiderman wants a Web that's social, but that he didn't talk about how someone could build it. Speaking as a technologist, Hendler said the Web works reasonably well for single document texts or for finding sites based on single document texts, but it can't integrate information from multiple documents. For example, he said it would be great to create custom traffic reports from a variety of sources, such as cameras, databases and statistics. He does that on the Web now by organizing information from lots of sites. But he wants the Web to do it for him. As it stands now, the Web will never really get any better. Given the exponential growth of the Web, information retrieval technology needs to get an order of magnitude better just to keep the same level of quality, he said.

We need agents on the Web, Hendler said, and here's why:

  •   When he talks to his human travel agent, a conversation emerges. He specifies his needs, constraints and preferences shared communication. The agent replies with options, costs and additional needs autonomous behaviors. He then decides what to do, and they do much of it for him   performing tasks.
  •   The next time he consults with his travel agent, their previous conversations are remembered. The agent gets better at predicting his preferences learning is occurring.

Hendler said he wants the computer to go beyond text. Instead of just giving him the weather, he wants the computer to tell him what satellite images are available, when they'll be updated and how much they'll cost. He wants a Web of context and meaning.

Web ontology, Hendler said, is a step in the right direction. The Web has no major organizing principle; it allows anyone to author and publish, and that arrangement allows the Web to grow arbitrarily large. We need a distributed way, he said, to provide organizing terms and terminologies and deploy them on the Web. We need pages marked with machine-readable terms, term-to-term (schema to schema) machine-based translations, and the ability to extend the ontologies for each domain.

Hendler acknowledged this vision will make a lot of bad things worse. He's worried about who's collecting the information. He added he did a search on the number of cows in Texas and found a site that says all the cows have been replaced by aliens. But those people need to be able to play, too the Web, he said, still needs to be the Web.

Shneiderman characterized Hendler's vision as "wishful thinking." He reiterated his call for user studies. He said using the term agent in conjunction with the user interface is a problem, because people are different from machines. Hendler's analogy of the travel agent is an error, Shneiderman said. We need to focus on responsibility and trust, and trusting a computer agent doesn't substitute for human-human trust. Humans are responsible, he said; machines aren't.

Hendler responded that he doesn't want to mimic human behavior on the Web, any more than he wants to build an airplane that mimics the way birds fly. Humans represent a terrible model for the Web. He agrees with Shneiderman that you can't "trust" a machine the way you "trust" a human but he doesn't have any other term to use to describe the concept.

At this point, Hahn invited the audience to join in the discussion. In response to one comment, Shneiderman asked when agents violate copyright, who's responsible? Hendler said the idea is to let a number of different approaches co-evolve; the best survive. Shneiderman said he agrees with that, but language matters: when you use the term agent, it's misleading. Hendler responded that if Shneiderman will tell him there are no bad interfaces, he'll admit there are bad agents.

After another member of the audience noted that at some level, there's no disagreement between Shneiderman and Hendler, Hendler responded that we need to discuss how we move forward with the Web. He said he and Shneiderman differ on their approach to co-evolution. He believes the technology changes your ideas on how you use it, and that change changes the technology you use. Shneiderman said he agrees with the concept of co-evolution. But he also wants a balance in which 50% of the development funding is spent on human behavior and 50% on technology. Hendler said he thinks the real debate between him and Shneiderman is where the points are set to obtain the best trade-offs.

The discussion continued along these lines for some time. It's safe to say the issues will remain controversial as we continue to develop the Web. You can obtain more information on the debate from the ASIST Web site at www.asis.org.

How to Order


ASIST Home Page

American Society for Information Science and Technology
8555 16th Street, Suite 850, Silver Spring, Maryland 20910, USA
Tel. 301-495-0900, Fax: 301-495-0810 | E-mail:
asis@asis.org

Copyright 2001, American Society for Information Science and Technology