Knowing What's Relevant: Factors in Selecting Relevant Information to form a Knowledge Base

• In Event: Poster Session 08
  In Poster Session: PS 08 Topic - Attention, Learning, Memory

Abstract

As humans, we form our knowledge base in a variety of ways- through direct experiences, inductive reasoning, deductive reasoning, and the subject of this research- self-derivation. Knowledge extension through self-derivation occurs when one integrates “stem facts” acquired in one learning episode (e.g., “The United states buys most of its avocados from Mexico”) with stem facts acquired in another (e.g., “Avocados are also called alligator pears”) to generate new knowledge (“The United States buys most of its alligator pears from Mexico”). Large developmental differences in children’s self-derivation have been documented between 4- and 10-years of age (Bauer, Blue, Xu & Esposito 2016). Previous research suggests that one source of age-related changes is the ability to select relevant stem facts (Bauer & Larkina, 2017).
In the current study, we investigated the conditions that affect 8- and 12-year-old children’s abilities to select relevant stem facts and self-derive new knowledge. We focused on whether the (a) temporal spacing between presentation of facts within a set, and (b) level of demand for selecting the relevant stem facts, influenced self-derivation. We presented 65 participants (M age = 9.94 years, range = 8.04-12.90, 32 girls, 39 8-year-olds) with 12 fact sets, 4 stem facts within each set. Two of the facts within a set could be integrated to self-derive a novel fact (integrable-facts) and the other two could not be integrated with any facts (non-integrable facts). Across participants, we manipulated presentation type: whether the facts were presented in a massed (all facts within a set presented consecutively) or distributed (facts within a set interleaved with other sets) order. Within participants, we manipulated demand type, whether the non-integrable facts within a set were high demand (same topic as integrable facts) or low demand (different topic as integrable facts), Figure 1. During encoding, facts were presented individually via pre-recorded audio and a corresponding image on the participant’s screen. After participants’ encoded the facts, we tested for self-derivation by asking them open-ended integration questions (e.g., “Where does the United States buy most of it alligator pears from?”) and followed up with forced-choice format if the participant answered inaccurately in open-ended.
We conducted a mixed effects ANOVA with presentation type (massed, distributed), demand (high, low), and age group (8,12), Figure 2. For open-ended testing, we found a main effect of age group F(1, 61) = 14.45, p < .001, such that 12-year-olds performed significantly better than 8-year-olds. There were no main effects of presentation type (p=.193) or demand (p=.222), nor any significant interactions (ps >.747). We found the same results pattern for the total score (open-ended + forced-choice) analysis. Thus, children were equally skilled at navigating mass vs. distributed presentation, whether they were required to select relevant facts from a small vs. a large set. Overall, this work contributes to understanding the different conditions that influence developmental changes in children’s learning and how those conditions influence their ability to productively form a knowledge base.

Authors

• Katherine Anne Lee, Emory University
  Presenting Author

• Melanie H. Hanft, Emory University
  Non-Presenting Author

• Jessica Ann Dugan, Emory University
  Non-Presenting Author

• Hilary E. Miller-Goldwater, Emory University
  Non-Presenting Author

• Patricia J Bauer, Emory University
  Non-Presenting Author