Proof in litigation is a decision making and inferential process. It is also a natural process. If proof in litigation is a natural process -- if, that is, proof in litigation is a real-world process--, it is a temporal, or dynamic, process.

If proof is a real-world temporal process, it is a process in which some form causality has some sway. This means that forensic proof is a process in which states and events at any point in the process can influence subsequent states and events in the process.

If all of this is true and if it is also true that proof is a probabilistic or stochastic process -- or if it is at least true that proof is known to us as a probabilistic process--, it follows that proof in litigation is -- or is known to us as -- a (natural) process with branching possibilities. (This means that any episode of proof in litigation can follow any one of a number of scenarios.) If there are branching possibilities or options at each step of the process of proof in litigation, the possible paths of proof in litigation have a tree-like structure.

Are the workings of this sort of process predictable? Is it possible to assess in an orderly way how decisions by participants in the process of proof may affect the workings of that process?

1. Assumptions and Stipulations

Assume that proof involves investigation.

Assume that a decision maker -- a trial lawyer, a prosecutor, a police officer, a private investigator, or a client -- is engaged in an investigation connected with litigation or with the possibility of litigation.

This decision maker takes investigative steps.

It is reasonable to assume that the decision maker's choice of investigative steps is influenced by the factual hypotheses and possibilities that the decision maker discerns and entertains.

Make this assumption.

Now consider Figure 1 (which I have borrowed from D. Schum & P. Tillers, "Marshalling Evidence in Adversary Litigation," 13 Cardozo L. Rev. 657 (1991)):

Figure 1 depicts investigative activity.

•Since investigation is a part of the process of proof, Figure 1 also depicts activity in proof. Furthermore, the structure of some non-investigative proof activity resembles the structure of investigative activity; stated otherwise, some proof activity has investigative characteristics. If so, Figure 1 also represents the structure of some non-investigative or nominally non-investigative proof activity.

The arcs running from any choice node on the left to the boxes and choice nodes on the right represent possible investigative steps or moves, possible steps or moves in investigation.

Investigative steps generate new evidence, new factual hypotheses (or "possibilities"), and new factual issues; that is, investigative steps generate new investigative "frames," which are represented by instances of T.

Each arc represents one possible investigative step or move.

Each sequence of arcs running from left to right represents a possible sequence of investigative steps or moves.

Each sequence of investigative steps or moves is an "investigative track," an "investigative path," or a "path" or "track" in investigation.

Each possible investigative track or path is a "scenario of investigative activity."

•Since investigation is part of proof, each "scenario of investigative activity" is also a "scenario of proof activity," or a possible sequence of steps or moves in the process or activity of proof.

Figure 1 suggests that factual hypotheses or possibilities influence investigative and proof activity. It suggests that factual hypotheses influence both (i) possible individual steps in investigation and proof and (ii) possible series of steps -- paths or tracks -- in forensic investigation and proof.

2. The Number of Factual Possibilities and the Number of Possible Steps and Tracks in Proof

There are two reasons why Figure 1 suggests that the possible number of possible investigative tracks in proof is enormous. First, Figure 1 shows that the number of arcs and, thus, investigative steps immediately following each choice node may exceed one (1). Second, Figure 1 depicts the investigative and proof process as a branching activity, one in which each downstream investigative step (i.e., each investigative step to the right of each choice node) produces a multiplicity of new hypotheses further downstream. If this is in fact the way that investigative hypotheses breed, the rate of increase of both the number of investigative hypotheses and the number of possible investigative tracks is exponential. Hence, if each choice node generates or confronts ten (10) new branches (and if each branch represents a possible investigative step), a decision maker who wishes to ponder all possible investigative paths open to her in four rounds of investigative activity would have to consider 10,000 combinations of investigative steps; i.e., her investigative activity could follow 10,000 distinct scenarios and she would have to consider 10,000 possible paths of investigative activity.

This example may greatly understate the number of possible tracks in investigation and proof. Recall that Figure 1 rests on the assumption that the number of possible investigative steps and investigative tracks is in part a function of the number of factual hypotheses or possibilities. This is a reasonable assumption. Recall that Figure 1 also assumes that the number of factual possibilities is in part a function of the nature and amount of evidence available to the decision maker. This is also a reasonable assumption. There is good reason to think, however, that numbers in the range of 10 sometimes wildly understate the number of factual possibilities that are likely to be suggested by even very small collections of evidentiary details.

3. The Influence of Evidentiary Details on the Formation and Multiplication of Factual Hypotheses in Litigation

Consider Figure 1A.

Figure 1A depicts several ways in which evidentiary details have a multiplier effect on the number of factual hypotheses in investigation and proof.

Consider Figure 1B:

Figure 1B is merely a restatement, a reformulation, of Figure 1A. The purpose of this reformulation is to highlight the following propositions:

1. evidentiary details influence and shape factual hypotheses

2. factual hypotheses influence and shape the possible course of investigation and proof in litigation

3. hence, evidentiary details influence and shape the possible course of investigation and proof in litigation

Let's try to get a sense of how many factual hypotheses collections of evidence can suggest or generate.

Consider the lower tier of boxes in Figure 1A and the equivalent array of four boxes at the extreme left in Figure 1B.

Those boxes represent various ways that evidentiary trifles can operate to generate, shape, and multiply factual hypotheses.

Those four boxes are also a crude representation of a theory of evidence marshaling operations that David Schum and I formulated quite some time ago. See P. Tillers & D. Schum, "A Theory of Preliminary Fact Investigation," 24 U.C. Davis L. Rev. 931 (1991).

The Schum-Tillers theory of evidence marshaling emphasizes the influence of evidentiary details on the formation, refinement, testing, and elimination of factual hypotheses. I cannot recapitulate that rather extended theory here and now. Suffice it to say that one of our principal conclusions is that quite small collections of evidentiary trifles can generate, support, or suggest very large numbers of factual hypotheses and possibilities. I am still convinced that this is true. Perhaps one illustration will be enough to convince you of this as well.

Assume that there are reports of the following four events:

XX states, "There was never a man as detestable as YY."

XX strikes YY's head.

YY strikes XX's head.

YY's skull cracks.

How many factual hypotheses or possibilities might the reports of these four events suggest or support?

Simplify this question by making the unrealistic assumption that the reports of these four events are perfectly credible and believable. Further simplify the question by assuming that our ruminations and reflections about these four reported events that I have described focus on chronological order.

Even if we make these two simplifying assumptions, it is fairly obvious that the four reported events that I have described can support or suggest a variety of factual hypotheses or possibilities in addition to the hypothesis consisting of a conjunction of the four reported events.

For example, the reports of the four events suggest following possible sequence of possible events (Sequence #1):

YY strikes XX in the head.

XX strikes YY's head.

YY's skull cracks.

XX states, "There was never a man as detestable as YY."

The following possible sequence of possible events (Sequence #2) is also supported by the reports of four events described above:

XX states, "There was never a man as detestable as YY."

XX strikes YY's head.

YY strikes XX's head.

YY's skull cracks.

The following possible sequence of possible events -- Sequence #3 -- is also supported or suggested by the previously-described reports of four events:

YY strikes XX in the head.

YY's skull cracks.

XX strikes YY's head.

XX states, "There was never a man as detestable as YY."

It is unnecessary to examine every possible sequence of the four reported events. It is enough to observe that Sequence #1, Sequence #2, and Sequence #3 are not identical factual hypotheses; they are different.

Furthermore, if I am not mistaken, where n = the number of events, the number of possible temporal sequences of those events is n! This means that if there are just five (5) reported events, the number of possible event sequences is 120. If there are just ten (10) reported events, the number of possible event chronologies is 604,800.

This is not all. If one is permitted to supplement reported events with one or more wholly hypothesized events, the number of possible event chronologies grows apace. For example, if there are two reported events -- "XX shoots YY" and "YY dies" -- the number of available event chronologies is more than 2! if one is free to hypothesize the additional event "XX yells, "I hate YY.'" Indeed, if such supplementation is permissible, the number of possible event chronology hypotheses is limited only or principally by the decision maker's imagination.

Finally, the amount of time between events can be very significant. For example, if there are two reported events -- "XX yells, 'I hate YY"" and "XX shoots YY" -- it is one thing to hypothesize that the event "XX yells, 'I hate YY" occurs 60 seconds before "XX shoots YY." It's quite another thing to hypothesize that the event "XX yells, "I hate YY'" takes place 60 years before "XX shoots YY."

If we assume that each factual possibility -- each possibility such as a possible event chronology -- suggests or generates a distinct possible investigative step, it is apparent that the number of possible investigative steps can be quite large. For example, if we have discover that we have a collection of reports of ten separate events, we already have reason to think that the number of possible investigative steps is at least 604,800 -- because, as we know, if there are ten undated events, those events can be arranged into at least 604,800 possible event chronologies.

Having these things in mind -- that is, keeping in mind how many factual possibilities can be suggested by even small collections of evidence --, it is also important to keep in mind that the total number of possible combinations of investigative steps cannot safely be determined simply by calculating the arithmetic sum of the numbers of factual hypotheses (such as event chronologies) that two or more separate collections of evidence suggest or support. The branching properties of a series of possible investigative steps have to be taken into account. For example, if there are two separate sets of reports of 10 distinct events and if each set of reports generates 604,800 possible event chronologies, it does not necessarily follow that the number of possible combinations of investigative steps is 1,209,600. As a general matter, the number of possible combinations of any set of events can greatly outnumber the number of such events. Trees such as those in Figure 3 describe how investigative events or steps may be combined. Those trees also assert that investigative activity is a sequential activity. The arcs running from left to right depict the possible connections between separate and distinct investigative steps. If one examines the properties of trees such as those shown in Figures 1 and 3 and if one views each arc as a possible investigative event or step, it follows that the number of possible combinations of investigative steps (and, more generally, "proof steps") equals xⁿ where x = the number of investigative options immediately following any choice node, and where n = the number of investigative steps in any chronological series of investigative steps. If this is the way things work, it is not difficult to conjure up situations in which the number of possible sequences or combinations of investigative steps practically boggles the mind. For example, if one wished to contemplate how things might transpire in investigation and proof in just three steps in the process of investigation and proof and if one assumes that each possible step in any possible sequence of three investigative steps generates a minimum of ten reported events, one would have to consider in excess of (6 x 10⁵)³ possible series of acts and events during investigation and proof -- which, by my calculations, equals 216,000,000,000,000,000 -- 216 quadrillion -- scenarios. That's a goodly number of possible scenarios of the course of investigative activity!

Now let's consider how certain legal rules affect the structure of problems of proof in litigation. Assume that you are a decision maker in the process of litigation and in the process of proof in litigation. Indeed, please make the horrifying assumption that you're a lawyer!

Now imagine that you are confronted with the need to make a decision that may affect the course of proof in a lawsuit that you are handling. For example, you're trying to decide whether you should make a motion for summary judgment or whether you should instead take a deposition and proceed with other formal discovery.

We have already seen that one class of reasons why you might have some difficulty in making such a decision: (i) the complexity of the evidence you already have, (ii) the difficulty of reaching drawing sound inferences from those evidence complexes about the numerous factual hypotheses and issues that you and other decision makers face or may face, and (iii) the difficulty of deciding which investigative or other steps {if any} should be taken to resolve or clarify some of the many unresolved questions and issues that may arise during the course litigation and proof in litigation. These are quite real and quite substantial difficulties. But they are not the only kinds of difficulties that can make a decision maker's job a daunting one.

A participant or decision maker in litigation must be aware that any step it takes may end up altering, perhaps to its detriment, the legal setting or framework in which the case or problem it is handling is embedded. A decision maker such as a lawyer must consider the possibility that its action may prompt a reaction from its opponent that will structure the subsequent phases of the lawsuit in a way that works to the initial actor's disadvantage. For example, lawyers often worry that they will "educate" their opponents by taking a step such as making a motion for summary judgment or a motion to dismiss; they worry that by taking such a step they will suggest claims and defenses to their opponents that otherwise would not or might not have occurred to their opponents. A participant in litigation, then, must decide whether it is worth running this kind of risk or a similar risk. But to make any such assessment, the decision maker must be able to make "reasonably good" predictions about an opponent's likely response or responses to the decision maker's possible choices.

The making of good predictions about such matters -- indeed, the making of any predictions about such matters -- seems to presuppose the ability of the decision maker to imagine the options, the courses of action, that are available to its opponent. This in itself may be hard to do; it may be quite difficult for a decision maker such as a lawyer just to imagine how an opposing lawyer might respond to a step such as the filing of a complaint or the making of a motion for summary judgment.

Possible sequences of actions and events in litigation might be represented as a series of linked boxes, or nodes, arranged in the form of a tree. For example, consider Figure 2:

The portrait of sequential decision making in litigation presented by Figure 2 is simple and straightforward. That's precisely the trouble: diagrams or graphs that simple cannot explain why a reasonably diligent attorney might have difficulty in imagining the possible consequences of her choices in litigation. They cannot do so because the choices, possibilities, and scenarios facing a decision maker in litigation are almost always more intricate and nuanced than a diagram or graph as simple as Figure 2 suggests.

One limitation of graphs having a structure like that in Figure 2 is that they may not adequately depict the full range of possible immediate responses to a procedural step such as a "complaint."

Another limitation of such simple graphs is that they may not sufficiently partition the initial steps or options that are available to a participant at any particular point in litigation.

Consider a lawyer who is trying to decide whether or not to file a complaint against a potential defendant -- and thereby officially commence a civil action. One way to describe the lawyer's immediate options is "the filing of a complaint" or "not filing a complaint." But there are other possible ways to describe her options with respect to a matter such as the filing of a complaint. Consider Expression 1.

Figure 2 asserts that a potential plaintiff -- "PP" -- has a choice between the filing of a complaint and non-filing of a complaint. Expression 1, by contrast, asserts that a complaint may take a variety of forms and, in particular, that it may contain any one of the following combinations of claims:

1. the defendant was negligent or committed battery or committed breach of contract

2. the defendant was negligent or committed battery

3. the defendant was negligent and committed battery and breach of contract

4. the defendant was negligent and committed breach of contract

5. the defendant committed battery and breach of contract

Expression 1 suggests that the PP's immediate options do not have to be described as being the filing or the non-filing of a complaint. They can instead be described as a choice among the non-filing of a complaint, the filing of a complaint with the claims stated in paragraph 1 above, the filing of a complaint with the claims stated in paragraph 2 above, the filing of a complaint with claims stated in paragraph 3 above, the filing of a complaint with the claims stated in paragraph 4 above, or the filing of a complaint with the claims stated in paragraph 5 above.

It is always possible (in principle) to describe any option or set of options in greater details, with greater granularity; it is always possible to partition the options more finely. For example, even Expression 1 may oversimplify the options available to a decision maker with respect to a complaint. For example, consider Expression 2.

Expression 2 is an even more granular representation of one of the options available to the plaintiff.

Details of the sort found in Expressions 1 and 2 may be important.

Consider Figure 3:

The tree shown in Figure 3 is a decision tree. I readily grant that Figure 3 is a corrupted decision tree; the notation in the graph or diagram violates several settled (and sensible) conventions. Nonetheless, Figure 3 is a decision tree in the following sense: it represents -- though imperfectly -- a decision maker's or analyst's state of mind about the possible courses of action and sequences of events that are harbored in a particular situation at a particular time.

In particular, the tree shown in Figure 3 represents, let us, say the thinking of Lousy Lawyer -- "LL" -- or Potential Plaintiff -- "PP" -- a legally knowledgeable person who is considering whether to start a lawsuit -- a civil action -- against David Defendant -- "DD" -- by filing a complaint. The tree in Figure 3 represents LL's or PP's view at time t₁ of how the situation or matter that interests her -- we can refer to that matter as a "case" or a "controversy" -- might evolve if she, LL, decides to file a complaint on behalf of her client.

Now this tree, which, as I said, represents our LL's view at time t₁ (or PP's) of how things stand and how they might evolve -- this tree, the tree in Figure 3, has several interesting features.

First, as you can see, the tree is rather bushy. This means that at time t₁ our decision maker -- LL or PP -- envisions a fairly large number of possible scenarios -- that is, she envisions a fairly large number of possible series of events in litigation.

Second, the tree in Figure 3 is a fairly detailed, or granular, representation of LL's or PP's view of her immediate options at the moment -- t₁ -- when she must make a decision or choice, and the tree provides a roughly equally granular description of (LL's view of) the possible responses of LL's opponent, DD, to LL's or PP's choices. This granularity of these descriptions is one reason why the tree you see in Figure 3 is rather bushy.

Third, while the tree in Figure 3 is rather bushy, it's entirely possible that nothing can be done about it. To wit: it's possible that the bushy tree in Figure 3 cannot or should not be pruned. That's because it's possible that the sorts of details displayed in Figure 3 have an important bearing on the decision that LL or PP faces at time t₁, the question being whether LL (or PP) should file a complaint. For example, it is entirely possible -- indeed I find this plausible -- that a lawyer such as LL in my hypothetical situation would both want and need to consider the possibility that her opponent might decide to file a counterclaim and it is entirely plausible that an actor such as LL would also both want and need to consider the possibility that a particular kind of complaint of hers might provoke her opponent to file a particular kind of counterclaim -- such as a counterclaim in antitrust or for defamation. The reason that she might want and need to think about this is that such a counterclaim might be one that LL, my hypothetical lawyer, particularly dreads and fears. (She might have reason to think that DD's chances of winning an antitrust claim would be excellent and she might have in mind that antitrust claims sometimes result in extraordinarily large damage awards.)

An important part of what I am saying that a useful or accurate tree sometimes justhas to be fairly dense and bushy if it is to accurately portray -- if it is to "capture" -- the factors that a decision maker such as a lawyer sometimes thinks are pertinent to the decision about litigation that she faces. But by emphasizing the necessity of a considerable amount of bushiness -- a considerable amount of detail -- in decision trees depicting problems of proof in litigation, I seem to be emphasizing -- or belaboring -- a familiar point: the sheer magnitude of the number of possible scenarios in the process of litigation and proof. I do not deny that this point about the multiplicity of such scenarios is part of the moral of my story or stories. But it is not the only moral of my stories. My stories suggest a distinct reason why it is hard to predict the course of investigation and proof in litigation (and why, by extension, it is hard to use automated, systematic, or orderly methods to assess decisions about the course of proof in litigation).

Consider Figure 3 once again. (See above.)

Suppose that the main litigation scenarios that a prospective plaintiff faces before starting a civil action are those shown in the (corrupted) decision tree shown in Figure 3. Suppose further that this prospective plaintiff {PP} most fears the scenario C & C_n > A & A_c & A_cd . This is primarily because he particularly fears A_cd , a counterclaim for defamation. However, A_cd aside, PP prefers to do C because he believes that doing C outweighs the disadvantages of all of the other responses he foresees his opponent might take.

Various things can happen that may diminish PP's fear of the event Acd. For example, PP can take step C_b rather than C_n, believing that C_b (a claim for battery) makes Acd (an answer containing a counterclaim for defamation) less likely. In general, when PP travels down the decision tree, things can happen that make some previously possible scenario seem much less probable than it did before. The moral of this part of my story is that any tree such as the tree in Figure 3 represents some one's state of mind, or picture of (a relevant sector of) the world, at a particular point in time -- for example, PP's state of mind at time t₁ . It follows, of course, that any such tree may not represent the state of mind of an actor or decision maker -- such as PP -- at a different time -- such as t₂.

Insert an new event after t₁ but before t₂. Let this event be "an interview with Walter Witness." Let this event be represented by the arc that branches downward from the box or node immediately to the right of Cn in Figure 4.

Once LL or PP takes this step -- once she or he has conducted the interview -- LL or PP no longer sees -- let me stipulate -- LL or PP no longer sees A_cd as a real possibility. That's why A_cd no longer even appears in Figure 4. Instead, as Figure 4 suggests, PP now sees only one likely response by his opponent, A_d, which is an answer containing only a denial of PP's claim, and no counterclaim. Perhaps she thinks this -- perhaps she now sees A_d as the only likely response by DD because, having acquired the evidence produced by the interview (see the node following the box representing the interview), PP now envisions the possibility of a claim for a violation of the Deceptive Practices Act -- C_p -- a claim that she did not envision before, she thinks that if she advances such a claim against he opponent DD, DD would not be in a position to do anything but to deny such a claim, and PP firmly expects and intends to amend her complaint to state a claim alleging solely a violation of the Deceptive Trade Practices Act. So if I amend and expand the tree shown in Figure 4 to reflect these assumptions and beliefs and events, we may get the tree shown in Figure 5:

What has happened? As Figure 6 suggests, one branch of the original tree has been completely cut off and an entirely new branch has been added.

The result is that the possible responses to C_n are no longer of interest and a new set of possible responses, those to C_p, are now PP's concern.

If we make some assumptions about LL's or PP's probability estimates, the effects of the events that I have described are rather dramatic. It's entirely possible that at time t₁ LL (or PP) thought that the investigation and proof scenario C_n > A_d & A_cd was the most likely and that it would therefore probably eventuate in T{C_n & A_cd}, a trial of PP's negligence claim and DD's counterclaim for defamation. This scenario is depicted by the bold lines in Figure 7.

However, once LL has conducted the interview I described earlier, that is, some time after t₁ , she no longer thinks the scenario shown in Figure 7 is a real possibility. She now thinks that the scenario depicted by the bold lines in Figure 8 is the way that things are most likely to work out.

Hence, instead of expecting a trial of the PP's claim in negligence and DD's counterclaim for defamation, she now expects a trial of her client's claim that DD engaged in deceptive practices, not at all what she anticipated when the first began thinking about the possibility of starting a lawsuit on PP's behalf against DD.

Observe that LL's or (PP's) revision of the decision tree is more than a revision. It is a reconstitution; i.e., except for the choice of C and not-C, LL's or PP's "revision" of the decision tree and of the decision problem she faces is a reconstruction rather than a revision. Observe, furthermore, that PP's reconstitution or reconstruction of the decision tree and of his or her the decision problem occurs as a result of new evidence. Observe, finally, that it might not have been possible for LL or PP at time t₁ to anticipate the reconceptualization of the decision problem depicted in Figures 6 and 8. My story suggests that until LL conducted the interview, LL and PP were in no position to anticipate the possibility of C_p, the possibility of a claim for deceptive practices.

1. Tentative Conclusions about the Limitations on the Foreseeability of the Course of Forensic Investigation and Proof

Like factual hypotheses, the branches of decision trees in litigation are extraordinarily sensitive to small details. In fact, part of the reason for the sensitivity of branching possibilities in litigation to evidentiary details is the extraordinary sensitivity of factual hypotheses to such trifles. The problem that the importance of evidentiary details in investigation, proof, and litigation presents for systematic analysis of decisions about proof in litigation is twofold. The first is that multiplicity of detail produces an even greater multiplicity of factual hypotheses. The multiplicity of factual hypotheses presents a problem in and of itself, so to speak -- that is, it creates an inferential problem, a problem for inferential reasoning. However, to the extent that the multiplicity of factual hypotheses influences the course of investigation and proof, the multiplicity of factual hypotheses also presents a potential obstacle to orderly assessment of decisions about the dynamics of investigation and proof in litigation. The potential obstacle is the large number of paths that investigation and proof in litigation can take and the problem is to try to assess and manage them in some sort of sensible or reasonable or effective fashion. So the first general obstacle that evidentiary details present is sheer multiplicity of the factual possibilities and the investigative and proof possibilities that such little trifles generate. This, I think, is a stereotypical version of the problem of complexity. However, the stories that I have told suggest that the problem of multiplicity is not the only obstacle that is created by the role of evidentiary details in the process of investigation and proof.

My examples suggest that the decisions one makes in litigation are extraordinarily sensitive to evidentiary details, that whether one decides to proceed one way or some other way depends on the fine evidentiary details that one has in one's possession. One difficulty that this poses for automated, systematic, and (even) orderly analysis of decisions about proof in litigation is that it is very difficult to predict the evidentiary details that one will acquire if one takes this or that step in litigation. I think there is a sharp difference between evidence and information. When I think of information, I think of matters that I know will help me resolve a given question or hypothesis. Evidence in litigation (particularly during investigative phases of the process of proof) often has a different flavor; it tends as much to generate hypotheses and questions as to resolve them. And when this happens, the devil is often in the details; that is, the nature of the hypotheses that evidence suggests depends on the fine details or characteristics of that evidence. And such fine details, it seems to me, are particularly hard to predict or anticipate before one actually has them in hand.

The third problem that the role of evidentiary detail presents is related to the second. The difficulty is that the hypotheses that one has about the possible course of proof in litigation are often rather diffuse. I think that a large part of the reason for this is that one cannot formulate many factual hypotheses with any degree of specificity -- and, thus, neither can one formulate possible investigative and proof strategies with any degree of specificity -- until one actually has in hand the evidentiary details that will come one's way as investigation and proof proceeds. So there is a real sense in which investigation and proof -- particularly investigation -- are a kind of blind man's buff, a kind of short in the dark. To some extent at least. The real question may be: to what extent, precisely?

2. Open Questions

Have I demonstrated that orderly or systematic techniques for assessing decisions in litigation and about proof in litigation are useless?

Far from it.

To a significant extent my argument has emphasized the multiplicity and the consequent "bushiness" of the branching possibilities in investigation, proof, and litigation. But any argument that this sort of multiplicity produces unmanageable complexity must take at least take into account a variety of pertinent considerations.

Some of these considerations are relatively mundane. For example, I have stressed the capacity of evidentiary details to multiply factual hypotheses and the possible tracks that litigation and proof in litigation can take. But it is possible that this thesis holds the greatest amount of water in relation to the exploratory and preliminary stages of the process of proof in litigation. There is some reason to think that as litigation progresses and matures the number of issues in play narrows and that the number of possible tracks that litigation and proof in litigation can take also greatly narrows.

But there are more general considerations that need to be weighed if one is to wrest any conclusions or hypotheses from my argument about the prospects for systematic analysis of proof in litigation. One such general consideration is the question of the relationship between complexity and intelligibility. It is true that complexity presents obstacles to computability. And it is possible that if any of our attempts to assess decisions about proof in litigation are ever to do any good, we will have to be able to calculate the probabilities of alternative scenarios in litigation. But I suspect that some of you know far better than some law teachers do that the mere magnitude of all of the possible consequences or effects that can flow from any action or event does not itself demonstrate that the possible effects of such an action or event are both indiscernible or unpredictable. If that were the case, no future events in this grand cosmos would be either discernible or predictable. The entire question (or almost the entire question), I think, is whether we can devise effective strategies for simplifying problems of proof in litigation. My argument was not designed to show that we cannot do so. My argument was designed to show the nature of, as it were, the causes or sources of some of the complexities and difficulties that participants in the process of proof in litigation face. Only if those complexities and difficulties are properly understood is it possible to try to devise strategies for deciding when certain kinds of complexities, possibilities, details, nuances, and scenarios can be ignored with relative or acceptable safety.

Having said this, I must add that the development of strategies for simplifying complex decision problems concerning proof in litigation is no simple task. Insofar as litigation and proof in litigation are concerned, it is very often true that the devil is in the details. In the face of this difficulty, one might surrender to the temptation to say that systematic procedures for assessing problems of proof in litigation are nothing more (and nothing less) than heuristic devices or procedures, devices that decision makers should avail themselves of only if and only to the extent that such procedures seem to them to clarify and order their own intuitions and insights, and otherwise not. This approach merits further study. However, I also do not quite want to leave it there -- with this emphasis on heuristics. I'll tell you why: some talk about the heuristic value of this or that theory or method sounds like a cheap and easy way out of a tough situation. I do not for a moment believe that the assessment of problems of proof in litigation can ever be fully automated; human judgment must and will play a crucial role in the process. But it does not necessarily follow that it is impossible to develop systematic procedures and routines that will help human beings more effectively manage and assess complex possibilities and probabilities that they would otherwise manage and assess much less efficiently and effectively. I would very much like to hear your ideas about what those sorts of devices and procedures might be.

I am not suggesting that simple enumeration one can determine how many factual hypotheses are supported by a collection of evidence. The numbers and calculations in the text are only designed to give a sense of how large a number of factual possibilities and hypotheses might be suggested or supported by collections of evidence of various sizes. Keep in mind that the calculation in the text take into account only one method of evidence marshaling. By the same token, some methods of evidence marshaling tend to reduce the number of factual possibilities and hypotheses.

I suspect that a good many lay people tend to think that the course of a lawsuit is pretty much set in stone. Some lay observers may also believe that a judge or a jury is both permitted and expected to apply any body of law that seems applicable to the case or controversy at hand. But that is generally not the way that the American system of litigation works: while there may be some steps in litigation that may have to be taken in every case (-- this is more likely to be true in criminal litigation than in civil litigation --), the choices of actors such as lawyers have a great deal of influence on the course that any particular lawsuit actually takes, and this influence extends to the kinds of substantive and procedural rules that will apply in a particular lawsuit.

As you can see, in Figure 4 I have -- of necessity -- eliminated the antiphonal quality of Figure 3, one in which circles represented choices facing PP, squares represented choices facing PP's opponent, and every circle was followed by a square. Now -- in Figure 4 and thereafter -- all choice nodes take the form of squares and they represent choices facing either PP or PP's opponent (who, let me stipulate, is David Defendant, or "DD" for short). In a sense, of course, this solution is not very elegant, for any number of reasons. For example, to mix circles and squares may seem akin to mixing apples and oranges -- because a decision tree, I assume, is supposed to represent the choices or options available to the decision maker (and I am assuming that the decision maker whose options I am trying to describe and assess is LL or PP, not DD). However, I a bit reluctant to adopt a convention that would make me portray the options available to DD as just possible alternative states of nature. This is partly because I have the sense that if I did that, consistency would require that I have alternating circles and boxes, and I'm not entirely sure I'm up to that job. So I hope you'll forgive me for the inelegance of some parts of my diagrams and graphs. Some of you folks can probably easily find some elegant solutions to my graphing problems. If so, I would certainly like to hear your suggestions!

 

Tillers' Home Page