Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Value-Focused Thinking: a chapter-by-chapter summary

12 Post author: Vaniver 23 January 2013 02:39AM

This is a chapter-by-chapter summary of Value-Focusing Thinking by Ralph Keeney. The hope of this summary is to present most of the value of reading the book in a tiny fraction of the space. Reading the original chapters will provide additional elaboration, examples, and secondary concepts, but unlike the textbooks I've reviewed before only those interested in learning more should need to read the full chapters.

I'll state my basic impression of the whole book up front: it is a very useful book for the 'soft half' of decision analysis, by which I mean framing problems, understanding objectives, and interacting with humans. For a more general and individual-focused introduction to decision analysis, I recommend Smart Choices (which Keeney was a coauthor on); VFT appears primarily targeted at facilitators and contains much focused material not in Smart Choices. I will not suggest targeted reading for particular readers, as Keeney does that well.

The first two paragraphs of the preface seem worth quoting in full:

Many books have been written about decisionmaking. They tell us how to solve decision problems. They do not tell us how to identify potential decision opportunities. They tell us how to analyze alternatives to choose the best one. They do not tell us how to create alternatives. They tell us how to evaluate alternatives given some quantitative objective function. They do not tell us how to articulate the qualitative objectives on which any appraisal of alternatives must rest. This book is different. It does what the others do not.

Almost all of the literature on decisionmaking concerns what to do after the crucial activities of identifying the decision problem, creating alternatives, and specifying objectives. But where do these decision problems, alternatives, and objectives come from? This book describes and illustrates the creative processes that you should follow to identify your decision problems, create alternatives, and articulate your objectives. These prescriptions are quite different from the way people typically pursue these activities.

1. Thinking about Values

Most standard decision analysis follows the path outlined in my previous sequence. The trouble with this approach is that it's reactive, assuming that you are already faced by a decision problem with known alternatives and preferences, and gives little advice on how to determine preferences between those alternatives or how to discover new alternatives, or how to see an opportunity to act where you did not see one before and behave pro-actively.

Value-Focused Thinking (VFT) puts values at the center of decision-making. Having explicit values makes it easier to rank existing alternatives, generate new alternatives, communicate and negotiate, and identify new decision opportunities to pursue. A decision problem is when an event appears and you must choose how to respond; a decision opportunity is when you actively decide to shift away from the status quo.

2. The Framework of Value-Focused Thinking

A decision frame consists of two parts: the decision context and the fundamental objectives. The decision context is the set of appropriate alternatives available to the decision-maker, and the fundamental objectives make clear what consequences of the decision are important enough to drive decision-making. Good decision-making is made much easier when the two are appropriately matched: the context for a fundamental objective should include all actions the decision-maker could take to affect that objective.

Objectives can generally be classified as means or ends objectives. It's important to focus on ends objectives to allow full creativity in crafting alternatives. One frame's ends objective is another frame's means objective, and so it is often natural to nest frames. Keeney gives the example of the EPA managing carbon monoxide pollution: they might want to minimize CO emissions, to minimize CO concentrations, to minimize population exposure to CO, or to minimize health effects due to CO. Each broadening of the fundamental objective should be accompanied by a broadening of the decision context, as the EPA has more tools to minimize health effects than they do to minimize CO emissions, though one of those tools is minimizing CO emissions. 

As we continue to extend the decision frame to the decision context of an individual's or an organization's entire existence, our ends objectives become strategic objectives. While decisions are rarely made at the level of the strategic decision context, having explicit and consistent strategic objectives can be a very helpful guide to determining the objective for more limited contexts, where most decisions are made.

3. Identifying and Structuring Objectives

This is a detailed and heavily technical chapter designed to teach facilitators how to identify and structure objectives.

To begin, create a list of objectives. When eliciting from multiple people, ask each to identify objectives separately, to prevent anchoring on the ideas from the first presenters.

There are many devices a facilitator can use to stimulate ideas. They might ask for a wish list, where they imagine how they would rank alternatives if constraints were discarded; for alternatives that are particularly good or particularly bad, and what makes them good or bad; for what problems or shortcomings the status quo has, and why these are problems; for what consequences might determine the desirability of alternatives; for the objectives that they think other stakeholders might have; or many others.

Once the list seems somewhat complete, the objective should be separated into means and ends by asking "Why is this objective important?" If the objective is important because if promotes another objective, it is a means objective; if it is important for its own sake or because it supports a strategic objective, it is a candidate for an ends objective. Sometimes, objectives will be specifications of each other: it is important to minimize child casualties in automobile accidents because it is important to minimize human casualties in automobile accidents.

This leads to two structures: a hierarchy of fundamental objectives and a networks of means-ends objectives. The objective hierarchy serves to clarify the objectives, and the objectives network serves to clarify how the alternatives might affect the objectives. Since the network is built with alternatives in mind, it should only include means under the realistic control of the decision-maker, even though there will be many other causal inputs to how well the objectives are accomplished. During this process, it is common to notice holes in the structures and discover missing objectives.

Keeney lists nine values that are desirable for fundamental objectives, but I will focus on the two most critical: the fundamental objective should be essential, in that the alternatives under consideration impact it, and controllable, in that all of the tools the decision-maker has to influence the objective are within the decision context. This is a balancing act: narrower objectives, like means objectives, will be controllable but not essential, and broader objectives, like strategic objectives, will be essential but not controllable. When determining what to eat for dinner, your quality of life is essential, but not controllable- most of your tools for affecting your quality of life don't deal with the problem of planning your dinner.

4. Measuring the Achievement of Objectives

Keeney uses the word attribute to refer to a measure of how well an objective was achieved. There are sometimes natural attributes- like measuring annual profit in millions of dollars per year- but it may be necessary to construct an attribute or find a proxy to measure.

The choice of measurement is important, as it implies value judgments about the objective. Keeney gives the example of measuring the objective "minimize the loss of life." One possible measure is the number of fatalities, which implies that preventing the death of a ten year old and an eighty year old are equally valuable. Another possible measure is the number of expected years lost, which would be 66 and 6, respectively, implying that preventing the death of the child is eleven times as important as preventing the death of the elder.

5. Quantifying Objectives with a Value Model

Once attributes to measure objectives have been selected, we can apply the standard tools of decision analysis, like VNM utility functions. Keeney describes several properties which may hold for joint utility functions on the set of attributes, and what the implies for the aggregation of those utility functions. In particular, for a linear combination of utility functions to be reasonable the preferences for individual attributes must be additively independent, that is, expressed in terms of their marginal probability distributions, rather than their joint probability distribution (see Fishburn).

Keeney also discusses the importance of unit conversion in tradeoffs between outcomes. It is not meaningful to saytthat minimizing casualties is five times as important as minimizing costs, but it is meaningful to say that reducing costs by five million dollars is as desirable as reducing casualties by one. (When utilities are nonlinear, it is still important to preserve the conversion from measured units to utilities when discussing relative importance.)

6. Uncovering Hidden Objectives

Measuring attributes can provide insight into values by clarifying them and making them more precise. English phrases generally support more interpretations than quantitative measurements.

Many objectives are counterproductive to discuss publicly, and so they may be hidden by design, rather than by ignorance. It is still worthwhile to privately acknowledge them and make sure the objectives are clearly separated.

When independence assumptions are violated, this generally means a fundamental objective has been overlooked or means objectives are being used as fundamental objectives. If it is difficult to assess the tradeoffs between fundamental objectives, that suggests that at least one of the objectives may be poorly defined or unclear.

7. Creating Alternatives for a Single Decisionmaker

Clarity in objectives aids in choosing between known alternatives, but much of the value from careful decision-making comes from the discovery of novel, superior alternatives. This chapter details methods to use the stated objectives to create alternatives.

One method is to consider alternatives that maximize each objective separately, and then alternatives that maximize pairwise combinations of objectives. This will raise more of the potential solution space to attention.

Specification of objectives aids in the development of alternatives. Keeney gives the example of considering adults and children separately when seeking to reduce the number of deaths and injuries in automobile accidents: different classes of alternatives will impact those classes separately, and without considering them separately alternatives only appropriate to one group might not be noticed.

The means-ends objective network also stimulates the creation of alternatives. It may also be useful to consider what could be done in the current decision context to advance strategic values. Focusing on particularly good consequences, and then figuring out how to attain those consequences, may be superior to focusing on how first.

Considering general solutions (i.e. properties of solutions), rather than individual feasible solutions, makes it easier to focus on parts of a large solution separately. It is also often the case that a combination of solutions will result in a superior solution, which VFT makes easier to find.

In many cases, the real alternatives under consideration now are the processes which will determine future action. This is not always obvious, and acknowledging it can lead to superior solutions.

At the end of the decision-making process, it is worthwhile to step back and reconsider the alternative about to be chosen. What reference class for good solutions is appropriate, now that you know what a good solution looks like? Is that the class that you have been considering?

8. Creating Alternatives for Multiple Decisionmakers

When making decisions for multiple stakeholders- either in a negotiation where all stakeholders must assent for an action to be taken or in a dictation where one stakeholder decides- the process by which alternatives are considered is likely to be important and help determine stakeholder satisfaction with the decision.

Eliciting values from other stakeholders is an important step in pleasing them, and asking them for alternatives makes use of their creativity.

When navigating altruistic situations, where the values of others matter strongly to you, it is often useful to have a policy of honestly providing private preferences and then explicitly aggregating them. Keeney gives the example of a couple choosing where to go to dinner: one may suggest a restaurant which they think will please the other, which the other accepts, thinking that will please the first. Both, in fact, preferred another alternative, which they missed out on because they did not communicate effectively.

Negotiation can be made much more efficient with VFT; knowing your preferences and the other party's preferences allows you to find the efficient frontier and win-win tradeoffs.

9. Identifying Decision Opportunities

Many alternatives which you could select to advance your values that are not obvious; many times, it is not even obvious that you should be looking for them. Devoting regular time to looking for and developing your ability to create new opportunities should pay large dividends.

New opportunities and new alternatives are complementary; they often take the form "I wish that..." or "We should...", and both suggest an area to direct focused attention.

The creation of strategic objectives can be done at multiple granularities, from a list of one executive's objectives to a quantitative value model with input from the entire organization. (The Hansonian analog for individuals is worth considering.) What level of description is worthwhile is a decision opportunity of its own.

While they should change only rarely, revisiting strategic objectives periodically- perhaps once a month- is a useful time to correct course and devote time to discovering decision opportunities.

Explicit resource inventories can help discover new opportunities, alternatives, and upcoming problems which can be averted.

As well, once resources have been committed to some plan, check for ways to get additional value for little additional cost. Business trips often present opportunities for additional experiences or connections at much lower cost than they would be on their own.

Monitoring achievement of objectives both allows feedback of how well current plans are achieving those objectives and keeps those objectives available in your mind.

Establishing a process by which others can present you with decision opportunities allows you to make the most of their creativity and expertise. Keeney gives the example of asking someone whose judgment you respect, "What should I be achieving or doing that I am not?" He adds that it often takes some prodding to get revealing responses, so prod.

Keeney also recommends using VFT to do empathetic negotiation. Proposals designed to also appeal to the other party more than the status quo (win-win)  will be much more warmly received than proposals designed to maximize your objectives.

Many objectives cannot be satisfied through your sole action; dating is perhaps the most obvious example. Those objectives tend to be achieved by being in the right place at the right time, and so the primary way to advance those objectives is to be in the right place more often and at better times. Creating the right place and selecting the right time are both decision opportunities well worth considering.

VFT is also ideally suited to situations where you have no idea what to do. By methodically seeking to determine your objectives, you can uncover enough about the situation to see clearly and act decisively.

10. Insights for the Decisionmaking Process

A deep and thorough understanding of the values inherent in a decision situation can provide important insights for all aspects of decision-making, and these insights make it possible to achieve much better consequences from the decisions we face.

VFT contributes to decision-making in six synergistic ways:

  1. Guiding Information Collection: Discovering values early in the process helps inform all other information collection. If values are unclear, they should be clarified; if something is unimportant, effort should not be wasted on it.
  2. Evaluating Alternatives: With clear and consistent objectives, VFT-powered analysis of objectives is often better than unsophisticated analysis.
  3. Interconnecting Decisions: Strategic objectives, fundamental objectives, attributes, and value tradeoffs are all likely to be useful in multiple decision contexts. By adopting the same values across the board, a decision-maker may be able to make better tradeoffs and satisfy their values better or at lower cost.
  4. Improving Communication: Clear and coherent values both improve thinking and communication. Multiple decision-makers can negotiate more effectively and act in concert with explicit and defined objectives.
  5. Facilitating Involvement in Multiple-Stakeholder Decisions: VFT provides a natural way to incorporate the preferences of other stakeholders into the decision-making process, often improving both their satisfaction with the process and the quality of the solution for all involved.
  6. Guiding Strategic Thinking: The strategic objectives of a decision-making activity, which is much easier when those objectives are clearly stated and quantified.
Developing a solid model of strategic objectives is particularly valuable because they have such broad impact and are unlikely to change, and so an investment now will repay in superior decisions for months and years with little maintenance required.

11-13. Part Four- Applications

I do not summarize Part IV of the book because it consists of extended examples, which I have mostly omitted for sake of brevity. Chapter 13, detailing the application of VFT to Keeney's life, will be of interest to most readers. In particular, his lists of strategic and professional objectives are worth considering once you have attempted to build lists of your own.

Comments (13)

Comment author: MaoShan 24 January 2013 02:45:49AM *  4 points [-]

practicularly (adv.) in the most practical instance of a set of several instances.

This could be a handy word; I'll have to try using it in a sentence sometime.

Comment author: Vaniver 24 January 2013 04:23:25AM 3 points [-]

Unintentional, I fear! I've edited it to "particularly."

Comment author: MaoShan 25 January 2013 12:51:36AM 2 points [-]

A serendipitous discovery nonetheless. :)

Comment author: Dr_Manhattan 24 January 2013 11:37:24AM 3 points [-]

The Hansonian analog

Which is - link?

thanks!

Comment author: Vaniver 25 January 2013 12:06:23AM 3 points [-]

I had in mind treating individuals as institutions, with interlocking and often competing subcomponets. I'm not finding a link that would be helpful; does anyone else know of a good summary or introduction?

Comment author: Dr_Manhattan 25 January 2013 02:12:49PM 2 points [-]

Basically the "Multiple Self" theory?

Comment author: Vaniver 25 January 2013 04:03:31PM 1 point [-]

Yep. I seem to remember him adding additional claims and flavors, like trading between selves and homo hypocritus (although that may be more a separate theory than an offshoot), that would be worthwhile to consider when investigating hidden objectives.

Comment author: Dr_Manhattan 25 January 2013 04:07:34PM *  0 points [-]

homo hypocritus

Without seeing the original source sounds like Robert Trivers stuff (who showed evolutionary basis for self-deception). But I agree that stuff sounds more low-level than multiple selves.

Comment author: Giles 03 February 2013 04:03:11AM 1 point [-]

There's probably better stuff around, but it made me think of Hanson's comments in this thread:

http://lesswrong.com/lw/v4/which_parts_are_me/

Comment author: Vaniver 03 February 2013 05:24:47AM 0 points [-]

In particular, this one. Thanks for pointing those out!

Comment author: Giles 03 February 2013 04:03:49AM 0 points [-]

There's probably better stuff around, but it made me think of Hanson's comments in this thread:

http://lesswrong.com/lw/v4/which_parts_are_me/

Comment author: fiddlemath 23 January 2013 04:19:06AM 1 point [-]

VFT appears primarily targeted at facilitators and contains much focused material not in VFT

er?

Comment author: Vaniver 23 January 2013 04:31:42AM 1 point [-]

Typo. Thanks for catching it!