Is that Right? Active Experience is Applied to Determine Teaching Quality

• In Event: Diverse Knowledge, Diverse Cues: Implications for Learners’ Inferences and Actions

Abstract

There is an interplay between learning through self-generated evidence and teaching from others. Moreover, we have varied tools to learn from both sources. For example, children assume that teaching is efficient and complete, but also are able to generate their own evidence in independent contexts. When one source is insufficient, children recruit the other; learning more from instruction after they explore and fail (DeCaro & Rittle-Johnson, 2012), and exploring more when a teacher omits information (Gweon et al., 2014). However, solutions that work for one learner may not work for others and even reliable teachers can provide inaccurate information. Therefore, in these situations the quality of teaching will not be clear unless children test the information themselves to infer its quality. This work investigates whether children can infer a teacher’s solution is poor by implementing and exploring the solution through their own actions.

Sixty-two children (mean age = 8.80 years, range: 6.05 – 11.43, 33 female) participated in a novel virtual platforming game. Children were pedagogically shown how to navigate the game map to find a trophy. Then, children were instructed to play the game for seven minutes and try to win. Unknown to participants, the solution was incorrect as platforms were removed from the map; however, exploring the map would provide alternate solutions. Given that teaching is generally more likely to be complete, a theoretically optimal pattern would be to test the solution, then strongly increase exploration after failure. Behavioral coding indexed whether participants generated evidence of the solution’s insufficiency by measuring whether they jumped where the platforms were removed. Additional coding was performed to see how they allocated their time across imitative and exploratory behaviors.

As a group, children made the rational assumption that the teaching was complete, imitating more than exploring, t(61) = 9.78, p < .001. Children also reliably produced evidence that the solution was insufficient in the course of imitating, t(61) = 4.32, p < .001. Among those that generated evidence, their rate of exploration significantly predicted solving, Z = 3.16, p < .01, reflecting the necessity to switch to finding their own solution. On the other hand, exploration did not predict solving among those that did not generate evidence, and those that did not generate evidence did not solve the task at lower rates, p > .05. Rather, this group explored more relative to their peers, t(25) = 3.75, p < .001, and solved more quickly, t(11) = 5.88, p < .001, suggesting that this group was more eager to disregard teaching and explore more spontaneously.

Therefore, children first attempt to exploit teaching to solve more efficiently, and in this process test the solution and generate evidence that it is incorrect, succeeding more as they increase exploration. However, those that do not generate evidence are not merely failing, but using a different learning strategy, exploring more from the beginning. Future work will seek to identify differences between children that generated evidence and those that did not by measuring spontaneous exploration.

Authors

• Mia Radovanovic, University of Toronto
  Presenting Author

• Ece Yucer, University of Toronto
  Non-Presenting Author

• Denise Arefhaghi, University of Toronto
  Non-Presenting Author

• Christie Lai, University of Toronto
  Non-Presenting Author

• Jessica A Sommerville, University of Toronto
  Non-Presenting Author