Professor Andrew Howes

November 2007

My research interests are in adaptive interaction, that is in how people find new and better ways to achieve their goals. I study the limits that social and cognitive mechanisms impose on adaptation.

With David Peebles and Rick Cooper I am chair of the International Conference on Cognitive Modeling 2009 which will be held in Manchester between July 23rd and July 26th of 2009. I am doctoral consortium chair for HCI 2008.

Cognitively bounded rational analysis

In collaboration with Richard Lewis at the University of Michigan and Alonso Vera at NASA Ames Research Center, I am exploring a novel approach to explaining adaptive behaviour. The key claim is that a theory can be said to explain behaviour if the optimal adaptation permitted by the theory corresponds to the empirically observed asymptotic bound on human performance. The approach is an alternative to rational analysis (Anderson, 1990), with which Anderson emphasised the constraint imposed by the task and the environment only, and to simulation with cognitive architectures (e.g. EPIC, ACT-R), with which it has been difficult to explore the bounds on adaptation. Cognitively bounded rational analysis emphasises the need to reason about the implications of theories for the bounds on adaptation.

We have developed constraint satisfaction techniques to support inference about the implications of theories. For more information see Howes, Vera, Lewis (2007) [pdf].

Information-Requirements Grammar

With Alonso Vera, Richard Lewis, and Juliet Richardson I have argued that existing languages for representing knowledge for routine cognitive tasks(such as GOMS, UAN, and PDL) can fail either because they demand that task competence is described using serial position to determine temporal order (and they are therefore overly restrictive) or because they demand that partial orderings are specified with temporal dependencies and other logical relationships (and they are therefore under-constrained). We have proposed a theory, called Information-Requirements Grammar (IRG), of how higher-level task knowledge constrains adaptation. The theory formalises a hypothesis about how higher level task adaptation is constrained by the information requirements and resource demands of lower-level tasks. For more information see Howes, Lewis, Vera, Richardson (2005) [pdf].

Strategies for Guiding Interactive Search

Duncan Brumby and I have been investigating the strategies that people use to search web pages. One activity people engage in when using the web is estimating the likelihood that labelled links will lead to their goal. However, they must also decide which items to assess and how to assess them. There are a number of theoretical accounts of this behaviour. The accounts differ in whether it is assumed that people consider all of the items on a page prior to making a selection or whether they make a selection immediately following an assessment of a highly relevant item.

We have conducted a number of experiments designed to discriminate between these accounts. The experiments systematically manipulated the relevance of the target and distracter items, and the location of the target item within the set. The findings suggest that decisions are continually made about whether to select one of the assessed items immediately or whether to make further assessments. Each decision is sensitive to the estimated relevance of all of the items so far assessed, and not just to the most recent item. The findings also suggest that when a goal-relevant item is located participants sometimes choose to check the remaining items in the menu but are more likely to skip some of these items.

Duncan has presented his work at two workshops and has an HCI journal paper accepted. For more information please email Duncan: Brumby@cs.drexel.edu.