Decision making under uncertainty, multiple frames and incompatible goals
CSIRO CELS Laboratories, 27-28 April 2011
What we want to do. We want to explore alternative avenues to model the decision making of agent(s) under uncertainty, when uncertainty arises from the presence of multiple frames (contexts under which the decision problem can be perceived) or incompatible goals.
Standard modelling practise (in agent-based modelling or artificial intelligence, for example) employs either sets of conditional probabilities to drive agents behaviour or adaptive techniques in which agents ‘learn’ how to act depending on feedback from the environment. These approaches necessarily ignore deeper cognitive aspects (1-9). In particular they ignore aspects of uncertainty which do not arise from a lack of knowledge, but from set of competing frames (10-12). This is particularly important for decision making among multiple agents.
Recent work (13-16) has attempted to model situations in which humans violate standard rules of probability theory (6, 8, 9), by treating probabilities within a Hilbert-space representation. While the approach is interesting and shows it can overcome some limitations from traditional approaches, it has been criticised for being still too naïve from the perspective of cognitive psychology. The purpose of this workshop is to seek expertise from cognitive psychologists in the design of a similar approach and to see whether their interaction with physicist and modellers can provide an improvement on this approach.
How we want to do it. We will let a small team of cognitive psychologists, sociologists, modellers and physicists brainstorm on the issue for 2 days. Current work in Hilbert-space representation of decision making will provide the initial inspiration, but will not constrain the approach.
No advanced mathematical knowledge is expected. It will be the responsibility of the modellers and physicists to simplify this aspect by the use of purely geometrical language (as much as possible). Communication between parties with different background will require that all concepts are discussed in plain, jargon-free English by everyone.
We expect attendees will make a reasonable effort to familiarise with the basic ideas before the workshop. It may necessary for the team to consult some literature during the workshop as well.
What we hope to achieve. We don’t expect miracles. Hopefully we will develop a better understanding on how to proceed or what avenues are unlikely to be fruitful. It is also possible that the outcome can be condensed in a joint paper.
In the ideal scenario, we may develop the structure of a general algorithm which can be used to model human behaviour in a number of complex, context-dependent decision-making settings.
In the context of scientific advice to policy making, this approach may provide an unlikely avenue of communication between natural scientists and engineers on one side and social scientists on the other. A geometrical approach may allow natural scientists and engineers to model, and thus more easily accept, the views social scientists hold so dear. Similarly, the adoption of such a framework may provide confidence to social scientists that some important aspect of social theory can be considered in quantitative models, making them to relevant to the real world problems they address.
What you should do now. Please provide 1-2 references you believe can contribute to the discussion. Also please specify whether you believe the overall team should read the references before the workshop or whether reading them is discretionary. In the former case, please be realistic by providing only really relevant literature. You can help by highlighting which part(s) of the reference have to be read. You can send them to Fabio for upload to this page or e-mail it to the overall team.
Some comments from Tim. ..... I agree very much with the sentiments that are being expressed; indeterminancy of decisions until the act; and the importance of framing or how the agents construe the situation (see the Jones et al paper in this special edition - not essential reading just illustrates that our thinking is compatible; mental models are compiled in real time in resposne to specific situations).
My expectation would however be that the dimensionality of real world decision situations may limit the applicability of the appoach. But that should not at all stop us from trying!
I suspect that there is quite a lot that we can do to reduce the dimensionality of the Hilbert spaces you are talking about (if I have understood the term correctly, which is not at all certain!). I see two important sources of large dimensionality; attributes or characteristics of the individual agents (which would include their construal of the situation, their mental models etc); and the characteristics of the setting (the place, the people that are involved, how they construe the setting and each other's participation, their relationships, roles etc).
In some recent reading I may be starting to see patterns in how we might constrain the dimensions of characteristics of the individual - Iain, you probably have an even better handle on this literature than I but things like goal pursuits (Ryan et al, 1996); attributional style (Peterson et al, 1982); uncertainty orientation (Fabio you have a reference to this in your list), the need for cognition (Cacioppo and Petty, 1982); the need for closure (Kruglanski et al, 1993) and the personal need for structure (Neuberg and Newsom, 1993). There are probably others...
My hope would be that we could find a relatively small number of dimensions that would define sets of agents with particular propensities or probabilities for particular decisions or decision sequences.
Constraining the dimensionality of settings is far more problematic for me. For what I think you are trying to do it may just be simplest to deal with a very narrow class of setting; small groups that are working together on a constrained task (e.g. a management team developing a strategic plan or some such thing).
As an additional reference I attach a paper on frame-based modelling by Tony Starfield and co. I do not think this is essential reading for anyone. It might be worth your looking over the introductory sections. Although an ecological approach I have been attracted to this from a social perspective as I suspect it would be interesting to try and model setting contexts (e.g. workshop, planning meeting) as frames (with slots for roles that are filled with people who have particular attributes across the lists above as well as relationships and slots for decisions etc). We might also think about frames of action....I have not given this much detailed thought but perhaps frames reflecting different stages in pathways to action. In part I am suggesting this as an alternative (though I think entirely complimentary) modelling approach; I have been struck recently at how powerfully the modelling paradigm that is adopted pre-determines what will be discovered so suspect we would benefit from multiple modelling pathways.
- Cacioppo, JT and RE Petty, 1982. The need for cognition. Journal of Personality and Social Psychology, 42:116-131
- Kruglanski, AW et al., 1993. Motivational resistence and openness to persuasion in the presence and absence of prior information. Journal of Personality and Social Psychology, 65: 861-876
- Neuberg, SL and JT Newsom, 1993. Personal need for structure: Individual differenecs in the desire for simple structure. Journal of Personality and Social Psychology, 65: 113-131
- Peterson, RE et al. 1982. The attributional style questionnaire. Cognitive therapy and research, 6:287-300
- Ryan, RM, KM Sheldon, T. Kasser and EL Deci. 1996. All goals are not created equal. An organismic perspective on the nature of goals and their regulation. In: (Eds) Gollwitzer, PM and JA Bargh, The Psychology of Action.
9am Kirsty Kitto - 'Plain English' Introduction to Hilbert Space representations
10am Iain Walker - Cognitive/social issues a meaningful theory or modelling approach need to account for
Where & When
27-28 April 2011 - Banksia Room, CSIRO CELS Laboratories, Underwood Ave (cnr Brockway Rd),
Floreat WA, 6014; , (08 9333 6000, Reception); View Map
1. D. Dorner, The Logic Of Failure: Recognizing And Avoiding Error In Complex Situations (Metropolitan Books, New York, 1996), pp.
2. J. Duckitt, C. G. Sibley. (Psychology Press, 2009), vol. 20, pp. 98 - 109.
3. J. T. Jost, J. Glaser, A. W. Kruglanski, F. J. Sulloway, Psychological Bulletin 129, 339 (2003).
4. E. Moxnes, System Dynamics Review 16, 325 (2000).
5. R. Sorrentino, C. Roney, Uncertain Mind: Individual Differences in Facing the Unknown (Psychology Press, Philadelphia, 2000), pp.
6. K. Stanovich, Who is rational? : studies of individual differences in reasoning N. J. Mahwah, Ed. (Lawrence Erlbaum Associates, 1999), pp.
7. J. D. Sterman, Science 322, 532 (October 24, 2008, 2008).
8. A. Tversky, D. Kahneman, Science 185, 1124 (1974).
9. A. Tversky, D. Kahneman, Psychological Review 90, 293 (1983).
10. M. Brugnach, A. Dewulf, C. Pahl-Wostl, T. Taillieu, Ecology and Society 13, 30 (2008).
11. R. Cross, A. Parker, L. Prusak, S. P. Borgatti, Organizational Dynamics 30, 100 (2001).
12. K. D. Joshi, S. Saonee, S. Suprateek. (Elsevier Science Publishers B. V., 2007), vol. 43, pp. 322-335.
13. J. R. Busemeyer, E. Pothos, R. Franco, Psychological Review (In Press).
14. R. Franco, Journal of Mathematical Psychology 53, 415 (2009).
15. A. Lambert Mogiliansky, S. Zamir, H. Zwirn, Journal of Mathematical Psychology 53, 349 (2009).
16. V. Yukalov, D. Sornette. (2009), vol. 11, pp. 1073-1120.
Cacioppo, JT and RE Petty, 1982. The need for cognition. Journal of Personality and Social Psychology, 42:116-131
Kruglanski, AW et al., 1993. Motivational resistence and openness to persuasion in the presence and absence of prior information. Journal of Personality and Social Psychology, 65: 861-876
Neuberg, SL and JT Newsom, 1993. Personal need for structure: Individual differenecs in the desire for simple structure. Journal of Personality and Social Psychology, 65: 113-131
Peterson, RE et al. 1982. The attributional style questionnaire. Cognitive therapy and research, 6:287-300
Ryan, RM, KM Sheldon, T. Kasser and EL Deci. 1996. All goals are not created equal. An organismic perspective on the nature of goals and their regulation. In: (Eds) Gollwitzer, PM and JA Bargh, The Psychology of Action.
Where to stay
- Floreat Bed & Breakfast, 9 Alderbury Street, Floreat WA 6014, (08) 9383 7112. This has been used 3-4 years ago by some colleagues who found it very pleasant and cheap. It is some 10-15 minutes of pleasant walking distance from the labs and similar distance to 3-4 restaurants in Floreat and Wembley. Only Bus connection the city and beach. It is on my way to the office so I can pick you up if you stay here.
- Most people stay in Cottosloe, to enjoy the beach. This hotel is very popular . You will probably need a car to get to the labs, or take a train and walk a bit. The ocean view, beach walks and morning swim may make it worthwhile. Many restaurants are available close by.
- Other places are available in Cottosloe, possibly cheaper