Automating argument construction

Over the past five years the Artificial Intelligence Center at SRI has been developing a new technology to address the problem of automated information management within real- world contexts. The result of this work is a body of techniques for automated reasoning from evidence that we call evidential reasoning. The techniques are based upon the mathematics of belief functions developed by Dempster and Shafer and have been successfully applied to a variety of problems including computer vision, multisensor integration, and intelligence analysis.

We have developed both a formal basis and a framework for implementating automated reasoning systems based upon these techniques. Both the formal and practical approach can be divided into four parts: (1) specifying a set of distinct propositional spaces, (2) specifying the interrelationships among these spaces, (3) representing bodies of evidence as belief distributions, and (4) establishing paths of the bodies for evidence to move through these spaces by means of evidential operations, eventually converging on spaces where the target questions can be answered. These steps specify a means for arguing from multiple bodies of evidence toward a particular (probabilistic) conclusion. Argument construction is the process by which such evidential analyses are constructed and is the analogue of constructing proof trees in a logical context.

This technology features the ability to reason from uncertain, incomplete, and occasionally inaccurate information based upon seven evidential operations: fusion, discounting, translation, projection, summarization, interpretation, and gisting. These operation are theoretically sound but have intuitive appeal as well.

In implementing this formal approach, we have found that evidential arguments can be represented as graphs. To support the construction, modification, and interrogation of evidential arguments, we have developed Gister. Gister provides an interactive, menu-driven, graphical interface that allows these graphical structures to be easily manipulated.

Our goal is to provide effective automated aids to domain experts for argument construction. Gister represents our first attempt at such an aid.     

Nest predators,but not nest survival,differ between adjacent urban habitats

Growing human populations make it imperative for ecologists to identify strategies to conserve biodiversity in human-dominated landscapes, such as cities. Effects of urbanization on birds are particularly well-studied, but questions remain regarding the best conservation approaches. Debate about the relative utility of focusing conservation efforts on nature reserves versus developed lands has focused largely on comparing species abundance or presence, with few studies addressing underlying behavioral or demographic mechanisms. Here we evaluated differences in avian reproductive success in nature reserves and matrix habitats to test the assumption that nest predation is lower within areas protected from development. Specifically, we investigated 1) whether nest survival differed in replicated pairs of forest parks and residential neighborhoods and 2) whether differences in nest survival were associated with changes in which species most frequently depredated nests. From April–August 2007–2014, we monitored nests of two native birds, American robin (Turdus migratorius) and northern cardinal (Cardinalis cardinalis), and video-documented nest predators in paired forest-matrix habitats in the Columbus, Ohio metropolitan area. We found similar rates of nest survival in the two habitats for both robins (Χ²₁?=?0.715, p?=?0.398, n?=?741 nests) and cardinals (Χ²₁?=?0.926, p?=?0.336, n?=?1156 nests), but interactions between predators and prey differed. In particular, domestic cats (Felis catus) were over five times as likely to depredate cardinal nests in matrix habitats versus forest parks (Χ²₁?=?7.24, simulated p?=?0.010; n_forest?=?3, n_matrix?=?7). Our results suggest that at least in some circumstances, nest success of native birds may be equivalent between nature reserves and adjacent residential matrix habitats, and thus residential neighborhoods may contribute positively to bird conservation in urban landscapes.     

What shapes plant and animal diversity on urban golf courses?

Recent concern over increasing loss of biodiversity has prompted considerable interest in the role of urban green spaces as reservoirs of local biodiversity. This study assessed the diversity of three indicator taxa - plants, ants and birds - on golf courses spanning a wide range of environmental variation in terms of climate, elevation, course age, size and connectivity to native woodland. Species richness and community composition was further compared between contrasting on-course habitat types that reflect different management intensities. We identified a set of taxon-specific environmental correlates indicating an intricate interplay of landscape- and local-scale variables that affect local species diversity. Our results show that floristic diversity is positively associated with the amount of rainfall, whereas ant and bird diversity are related to local-scale factors, particularly the number of trees and the size of water features on a site. The amount of on-course native habitat was a strong predictor of plant and ant diversity and was also associated with the number of unique species at the site level; this reinforces the value of remnant habitat patches as local biodiversity reservoirs that represent mini hot-spots in an otherwise species-poor urban landscape. Community composition for all three taxa differed markedly between non-playing and playing areas, with boundary and remnant habitats generally having more diverse, species-rich communities. Our results suggest that local floral and faunal biodiversity on urban golf courses can be enhanced by creating woody non-playing areas and, especially, by preserving, restoring or expanding remnant habitats.     

Landscape urbanization threatens plant phylogenetic diversity in the Brazilian Atlantic Forest

Urbanization causes species loss around the world, but its effects on phylogenetic diversity are poorly known in tropical forests. Using a patch-landscape approach in an urbanizing region of the Brazilian Atlantic Forest, we tested whether the increase in landscape urbanization reduces plant species density, phylogenetic richness and divergence, and increases the relatedness among co-occurring individuals and species. We assessed plant responses to urbanization in adult (diameter at breast height?>?10 cm) and sapling communities (2.5–10 cm diameter) separately, as saplings are proxies of the future flora. We sampled 2860 woody plants belonging to 155 species in nine circular landscapes with urbanization level varying from 0% to 45%, and estimated the relatedness among the species that have increased and decreased in relative abundance in more urbanized landscapes (winner and losers, respectively). As expected, species density and phylogenetic richness decreased with the increase in urbanization. These responses were consistent for adult and sapling communities, suggesting a persistent loss of species and lineages in more urbanized landscapes. Contrary to our expectations, phylogenetic divergence and structure did not respond to urbanization, indicating that the more urbanized landscapes still retain much evolutionary history. However, because the relatedness among winners was greater than among losers, it is likely that the phylogenetic divergence gradually reduces and the relatedness increases, resulting in impoverished forests with uncertain ability to provide ecosystem services such as carbon storage and pest control. This environmental cost should be taken into account to align urban sprawl with biodiversity conservation.     

Quick,uncertain, and delayed adults: timing,sequencing and duration of youth-adult transitions in Canada

The transition from youth to adulthood in western societies has become more prolonged and complex as traditional societal norms have lost influence. Using retrospective data from a cohort of 43-year-old Canadians surveyed in 2010 (n?=?405), we mapped the timing, sequencing, and duration of ‘first stage’ youth-adult transitions (leaving home, finishing formal education, obtaining a full-time job) and ‘second stage’ youth-adult transitions (marriage, parenthood, home ownership). Latent profile analysis identified five distinct transition profiles – norm-setters, quick adults, uncertain adults, adult students, and delayed adults – with different amounts and timing of post-secondary educational investment playing an important profile-shaping role. Although youth-adult transitions have become less standardized and more individualized, especially for first stage transitions, these transitions are still quite structured and continue to be affected by gender and family socioeconomic status.     

Strategies and cues adolescents use to assess the age of an online stranger

A common risk among adolescents is sexual solicitation, in which an adolescent is asked to provide sexual information, engage in sexual talk, or in sexual activities. Although scholars increasingly address this topic from an intrapersonal perspective, there is little attention to factors of language use and message content. In two focus group studies, we investigated whether adolescent girls consider themselves capable of assessing whether an online stranger is an adult or a peer, the extent to which adolescent girls actually succeed in making this assessment, the strategies they apply to do so, and the content- and language-related cues focused on. Our findings suggest that most of the adolescent girls are confident in their ability to assess whether the stranger is a peer or an adult with possible sexual intentions. However, we also found that only 43% were able to correctly make this assessment. Most of the adolescents seem to apply the passive strategy of uncertainty reduction: They scan the profile page of the stranger, check contact information, and the profile picture. They may choose to apply the interactive strategy to find out more about the online stranger. Important content-related cues that alarm adolescent girls are: ignoring personal questions, showing an exaggerated amount of interest, acting as a friend, and being sexually oriented. Language cues mentioned related to word usage, abbreviations, sentence length, and tone.     

Early Segmentation Abilities in Preterm Infants

Infants born preterm have higher risks of developing linguistic deficits. Considering that the ability to segment words from fluent speech is crucial for lexical acquisition, Experiment 1 tested the ability of healthy extremely‐to‐late preterm infants to segment monosyllabic words at 6 months of postnatal age. Results establish basic segmentation skills in these infants. While we failed to find an effect of the degree of prematurity, this issue will need further exploration. Future studies will also have to specify the scope of these early segmentation skills, both in terms of the types of words segmented, the cues used to do so, and in terms of possible differences in performance between subgroups of preterm infants (e.g., based on gestational age or medical risks). Lastly, given that the preterm infants tested had a mean maturational age of 4 months, Experiment 2 explored monosyllabic segmentation in full‐term 4‐month olds. Infants succeeded at the task, providing the earliest developmental evidence of word segmentation in full‐term infants. These findings better specify the early trajectory of segmentation abilities in both full‐term and healthy, low‐risk preterm infants and support the proposal that prematurity might have a differential effect on the early acquisition of various linguistic levels.     

Negative (But Not Positive) Parenting Interacts with Infant Negative Affect to Predict Infant Approach: Evidence of Diathesis‐Stress

Temperament by parenting interactions may reflect that individuals with greater risk are more likely to experience negative outcomes in adverse contexts (diathesis‐stress) or that these individuals are more susceptible to contextual influences in a “for better or for worse” pattern (differential susceptibility). Although such interactions have been identified for a variety of child outcomes, prior research has not examined approach characteristics—excitement and approach toward pleasurable activities—in the first year of life. Therefore, this study investigated whether 6‐month maternal reported infant negative affect—a phenotypic marker of risk/susceptibility—interacted with 8‐month observed parenting behaviors (positive parenting, negative parenting) to predict 12‐month infant behavioral approach. Based on a sample of mothers and their infants (N = 150), results indicated that negative parenting was inversely associated with subsequent approach for infants with high, but not low, levels of negative affect. Similar results did not occur regarding positive parenting. These findings better fit a diathesis‐stress model rather than a differential susceptibility model. Implications and limitations of these findings are discussed.     

The Influence of Maternal Socialization on Infants’ Social Evaluation in Two Cultures

Although a growing body of research has explored the early development of social evaluation, no research has directly compared social evaluations of infants between different cultures. In addition, there has been little understanding regarding socialization's effects on this ability. The goal of this study was to expand on earlier findings on social evaluation in infants by investigating a broader sample from two cultures, and to explore the influence of maternal socialization on infants’ social evaluation. Using the violation of expectations and the preferential reaching paradigm, four groups aged 6‐, 9‐, 12‐, and 15–18 months and their mothers from Japan and the United States (159 dyads) were compared in terms of spontaneous social evaluations. Japanese and European American infants showed similar performance in dishabituation to the inconsistent behavior and in their reaching preference for prosocial over antisocial agents, indicating that the emergence of spontaneous social evaluation is not culture‐specific. Furthermore, our study provides a novel finding regarding the relationship between mothers’ socially evaluative speech and infants’ preference for prosocial over antisocial agents. These results suggest that the development of sociomoral understanding results from complicated interactions among evolutionary, cognitive, and social factors.