By Rassbach, Laura; Bradley, Elizabeth; Anderson, Ken M.
AI Magazine , Vol. 32, No. 2
Automating scientific reasoning is an important challenge to AI. An automated tool can do boring and repetitive reasoning, freeing experts to do more difficult and creative work. Indirectly, it can make explicit the knowledge and reasoning used by experts in the field. Finally, an automated tool can consider all possibilities, sometimes exploring scenarios that human experts may miss.
This article discusses automating reasoning for dating geological landforms. Dating landforms is similar to investigating a crime scene: from the information available on the surface, left behind by an unknown series of events, experts must deduce what happened in the past. In the example diagrammed in figure 1, subsurface rocks are exposed over time as the soil around them erodes. A geoscientist would be faced with the situation shown on the right of the figure; his task is to deduce the situation shown at the left, along with the processes that were at work and the timeline involved.
To accomplish this, a geoscientist first dates a set of rock samples from the present surface, then reasons backward to deduce what process affected the original landform. This is a difficult deduction: geological processes take place over an extremely long period of time, and evidence remaining today is scarce and noisy. Finally, experts in geological dating, like experts in any field, are only human, and can be biased in favor of one theory over another.
In the face of these problems, experts form an exhaustive list of possible hypotheses and consider the evidence for and against each one--much like the AI concept of argumentation. Our system to automate this reasoning, Calvin, uses the same argumentation process as experts, comparing the strength of the evidence for and against a set of hypotheses before coming to a conclusion. We collected knowledge about how isotope dating experts reason through interviews with several dozen geoscientists. Confidence is key in this kind of reasoning, not only in the quality of evidence, but also in the knowledge that is used to connect evidence to conclusion. Capturing these elements required a novel instantiation of confidence-based reasoning in an argumentation system. From these elements, Calvin produces arguments almost identical to the reasoning presented by human experts.
[FIGURE 1 OMITTED]
Calvin provides several contributions to AI and to the larger scientific community. Its rule base is an explicit representation of the knowledge of two dozen experts in landform dating. It incorporates a rich system of confidence that captures the reasoning of real scientists in a useful way. It is a fully implemented and deployed system--a surprisingly rare thing in the argumentation literature. Finally, it is a real tool that is in daily use by real scientists.
In the following section, we discuss the general problem of cosmogenic isotope dating, highlighting its challenges and the approach that experts take to solving it. Next, we describe how Calvin uses argumentation to automate that process, and finally, we discuss our results.
Cosmogenic Isotope Dating
Beginning from a set of samples collected from boulders on a landform, an isotope dating expert's goal is to determine the absolute age of that landform. This section summarizes how experts work, from sampling individual boulders to deducing an age for an entire landform.
The first step is to collect as many samples as possible from the landform. A set of at least five samples is best (Putkonen and Swanson 2003); five to ten samples is about the norm. Experts would prefer to collect far more samples, but often only a handful of boulders suitable for sampling are available. While collecting samples, the expert also makes qualitative field observations that are often crucial for interpreting initial dating results.
Once the expert has gathered a set of samples in the field, he brings them to a lab for dating. …