Search for: All records

While a vast collection of explainable AI (XAI) algorithms has been developed in recent years, they have been criticized for significant gaps with how humans produce and consume explanations. As a result, current XAI techniques are often found to be hard to use and lack effectiveness. In this work, we attempt to close these gaps by making AI explanations selective ---a fundamental property of human explanations---by selectively presenting a subset of model reasoning based on what aligns with the recipient's preferences. We propose a general framework for generating selective explanations by leveraging human input on a small dataset. This framework opens up a rich design space that accounts for different selectivity goals, types of input, and more. As a showcase, we use a decision-support task to explore selective explanations based on what the decision-maker would consider relevant to the decision task. We conducted two experimental studies to examine three paradigms based on our proposed framework: in Study 1, we ask the participants to provide critique-based or open-ended input to generate selective explanations (self-input). In Study 2, we show the participants selective explanations based on input from a panel of similar users (annotator input). Our experiments demonstrate the promise of selective explanations in reducing over-reliance on AI and improving collaborative decision making and subjective perceptions of the AI system, but also paint a nuanced picture that attributes some of these positive effects to the opportunity to provide one's own input to augment AI explanations. Overall, our work proposes a novel XAI framework inspired by human communication behaviors and demonstrates its potential to encourage future work to make AI explanations more human-compatible.

 

The Nonesuch Formation microbiota provide a window into ca. 1075 Ma life within the interior of ancient North America. The Nonesuch water body formed following the cessation of widespread volcanism within the Midcontinent Rift as the basin continued to subside. In northern Michigan and Wisconsin, USA, the Copper Harbor Conglomerate records terrestrial alluvial fan and fluvial plain environments that transitioned into subaqueous lacustrine deposition of the Nonesuch Formation. These units thin toward a paleotopographic high associated with the Brownstone Falls angular unconformity. Due to these “Brownstone Highlands,” we were able to explore the paleoenvironment laterally at different depths in contemporaneous deposits. Rock magnetic data constrain that when the lake was shallow, it was oxygenated as evidenced by an oxidized mineral assemblage. Oxygen levels were lower at greater depth—in the deepest portions of the water body, anoxic conditions are recorded. An intermediate facies in depth and redox between these endmembers preserves detrital magnetite and hematite, which can be present in high abundance due to the proximal volcanic highlands. This magnetic facies enabled the development of a paleomagnetic pole based on both detrital magnetite and hematite that constrains the paleolatitude of the lake to 7.1 ± 2.8°N. Sediments of the intermediate facies preserve exquisite organic-walled microfossils, with microfossils being less diverse to absent in the anoxic facies where amorphous organic matter is more likely to be preserved. The assemblage of cyanobacteria and eukaryotes (both photoautotrophs and heterotrophs) lived within the oxygenated waters of this tropical Mesoproterozoic water body. 

The paleogeography of Laurentia throughout the Neoproterozoic is critical for reconstructing global paleogeography due to its central position in the supercontinent Rodinia. We develop a new paleomagnetic pole from red siltstones and fine‐grained sandstones of the early Neoproterozoic Jacobsville Formation which is now constrained to be ca. 990 Ma in age. High‐resolution thermal demagnetization experiments resolve detrital remanent magnetizations held by hematite. These directions were reoriented within siltstone intraclasts and pass intraformational conglomerate tests—giving confidence that the magnetization is detrital and primary. An inclination‐corrected mean paleomagnetic pole position for the Jacobsville Formation indicates that Laurentia's motion slowed down significantly following the onset of the Grenvillian orogeny. Prior rapid plate motion associated with closure of the Unimos Ocean between 1,110 and 1,090 Ma transitioned to slow drift of Laurentia across the equator in the late Mesoproterozoic to early Neoproterozoic. We interpret the distinct position of this well‐dated pole from those in the Grenville orogen that have been assigned a similar age to indicate that the ages of the poles associated with the Grenville Loop likely need to be revised to be younger due to prolonged exhumation.