No CrossRef data available.
Article contents
Understanding social networks requires more than two dimensions
Published online by Cambridge University Press: 26 February 2014
Abstract
The proposed framework is insufficient to categorize and understand current evidence on decision making. There are some ambiguities in the questions asked that require additional distinctions between correctness and accuracy, decision making and learning, accuracy and confidence, and social influence and empowerment. Social learning techniques are not all the same: Behavior copying is quite different from theory passing. Sigmoidal acquisition curves are not unique to social learning and are often mistaken for other accelerating curves.
Information
- Type
- Open Peer Commentary
- Information
- Copyright
- Copyright © Cambridge University Press 2014
References
Beppu, A. & Griffiths, T. L. (2009) Iterated learning and the cultural ratchet. In: Proceedings of the 31st Annual Conference of the Cognitive Science Society. ed. Taatgen, N. A. & van Rijn, H., pp. 2089–94. Cognitive Science Society.Google Scholar
Busemeyer, J. R. & Johnson, J. G. (2006) Micro-process models of decision-making. In: Cambridge handbook of computational cognitive modeling, ed. Sun, R., pp. 302–21. Cambridge University Press.Google Scholar
Montrey, M. & Shultz, T. R. (2010) Evolution of social learning strategies. In: Proceedings of the Ninth IEEE International Conference on Development and Learning, Ann Arbor, MI, August 18-21, 2010, pp. 95–100. IEEE.Google Scholar
Reader, S. M. (2004) Distinguishing social and asocial learning using diffusion dynamics. Learning and Behavior
32(1):90–104.Google Scholar
You have
Access
Bentley et al. address the issue of how people make decisions in the realm of large-scale social media. They propose an interesting two-dimensional framework in which to classify major influences on decision making: an east–west axis reflecting the degree to which decisions are made individually versus socially influenced, and a north–south axis reflecting the degree of transparency of the payoffs and risks associated with particular decisions. This is a useful start, and Bentley et al. get a certain amount of mileage from considering decision making within this framework. However, it is already apparent to us that this framework is insufficient to categorize and understand current evidence on decision making.
The question being asked
An issue that overshadows the work is a lack of clarity as to what aspect of human decision making they want to approach. A model is best evaluated within the context of the question it is being used to answer. For example, two specific questions that seem relevant to the proposed framework are: (1) “How do social context and critical thinking interact in producing right or wrong thinking?” and (2) “What is the role of social engagement in decision making?” These questions are reasonable, but involve the use of the proposed model in different ways. Furthermore, neither can be answered using the framework as proposed: For case 1, the model lacks a notion of correctness; for case 2, the model is unclear about modes of social interaction. In the target article text, the authors do not identify a particular, precise question that their approach addresses. This issue derives, in part, from ambiguity about how certain aspects of the proposed model should be interpreted. We highlight three of these below.
Decision making or learning?
Despite their discussion of decision making in their introductory remarks, it is sometimes unclear which of these human processes Bentley et al. are applying their model to – decision making or learning. Although decision making and learning are related, they are distinct processes that cannot be lightly interchanged (Busemeyer & Johnson Reference Busemeyer, Johnson and Sun2006).
Accurate or confident perception?
The proposed north–south axis is described in terms of clarity of information. But what is the nature of this clarity? Are we concerned with whether the individual feels confident about their perception of risks and rewards, irrespective of whether their perception is correct; or are we concerned with whether the individual accurately perceives the risk/rewards? The two are not the same thing and may need to be separated into different axes.
Social influence or empowerment?
The social (east–west) axis characterizes an ambiguous social dimension. Does moving east involve other people having increasing sway over what the individual believes (social influence) or other people having increasing sway over whether an individual acts on their belief (social empowerment)? Each of these is valid, but they reflect quite different and important aspects of decision making in a social context.
Lumping together all social learning methods
Bentley et al. lump a variety of social learning techniques together under a high value of their J parameter. However, there is now evidence that cultural transmission, defined as theory passing, has sharply different characteristics from imitation (Montrey & Shultz Reference Montrey and Shultz2010). Imitation (defined as attempted copying of behavior) is a lossy way to transmit information and may quickly become outdated in rapidly changing environments. In contrast, cultural transmission via passing of theories from one agent to another builds on existing knowledge, creating a strong ratchet effect with very little backsliding. In the Montrey and Shultz study, there were three learning methods in an agent-based model on a fully occupied lattice: imitation, exploration, and theory passing. Each learning method was implemented as a variant of Bayesian learning. As in previous work (Beppu & Griffiths Reference Beppu, Griffiths, Taatgen and van Rijn2009), three alleles were compared: imitation alone (least adaptive), imitation plus exploration (moderately adaptive), and imitation plus exploration plus theory passing (most adaptive). Agents reproduced by cloning an offspring according to their own fitness.
Shape of acquisition curve is not definitive
The authors use the shape of acquisition curves to unequivocally identify underlying learning strategies. But this method is known to be difficult and often unreliable because the same shape can be consistent with more than one learning method. For example, individual learning typically produces a sigmoidal curve (Shultz Reference Shultz2003), a shape that Bentley et al. assume is a unique signature of social learning. Not only is the overall acquisition curve often sigmoidal (basically, a spurt connecting two plateaus), but with denser time sampling, numerous sigmoidal spurts can be found. Moreover, a number of different accelerating functions have been successfully fit to social learning data (Reader Reference Reader2004). Even if the underlying social learning curve is truly sigmoidal, it can be mistakenly viewed as accelerating if final data points are missed, or as decelerating if early data points are missed.
Closing remarks
Despite the concerns raised above, Bentley et al. meaningfully contribute on an important topic: how social, media, and internal milieus interact to inform human decision making. The problem is highly dimensional, and Bentley et al. have provided a framework on which future work can productively build.