Emotions are commonly thought to be beyond the pale of rational analysis, for they are subjective, may vary even with respect to the person experiencing the emotion, and may con ict with rational thought. In this paper, we develop the position that emotions can be the objects of argumentation, which we express by introducing emotion terms in emotional argumentation schemes. Thus, we can argue about whether or not, according to normative standards and available evidence, it is plausible that an individual had a particular emotion. This is particularly salient in legal cases, where decisions can depend on explicit arguments about emotional states.
Emotions are commonly thought to be beyond the pale of rational analysis. They
are subjective; the same person in the same context may have di erent emotional
responses to stimuli; a person's emotional response may con ict with rational
thought. Emotions are also thought to only serve in an adjunct role in
decisionmaking, by enhancing, moderating, or interfering with the persuasiveness of
reasoning in an argument [
In this paper, we brie y outline the legal context, computational analyses of emotions, and current research on emotions and argumentation. We then introduce our novel emotional argumentation schemes, where the key idea is that emotional terms can be central components in the schemes. These schemes model key parts of reasoning in the legal context and of the computational analysis of emotions. We use the schemes to model legal arguments that are relevant to legal cases. We close with some indications of future research.
As emotions are a widespread, salient experience of our lives and in our social
encounters, it is unremarkable that they are the subject of legal proceedings, where
human experience and behaviour is reasoned about and regulated. Considering
legal contexts bounds our discussion in three respects. First, there are explicit
arguments about emotions, so we need only be concerned with explicit
statements about emotions rather than their psychological or physical reality. Second,
the legal context is normative and truth determining ; judges and juries decide,
relative to a normative model of human emotional responses. This means that
though a party to a legal case may claim an emotional state as justi cation for an
action, the courts may decide otherwise based on arguments about evidence,
testimony, normative reasoning about emotional states, etc. Third, the arguments
we consider are about emotional states after the fact, for we are not considering
emotions engendered during the court proceedings. This means that scienti c
indicators of the embodiment of the emotion, e.g. MRI brain scans along with
other physiological measurements, are not relevant to our discussion. While we
acknowledge theories bearing on the embodiment of mind and emotion [
Emotions in law appear in a variety of ways. In common law, among the causes of
action we nd intentional or negligent in iction of emotional distress and sexual
harassment, which have emotional referents. Over the course of litigation, there
will be arguments as to whether distress was caused, the extent of distress, along
with supporting evidence or expert testimony. In hate crimes, the emotional
disposition of the perpetrator, whether the perpetrator felt hate towards the victim,
may be subject to argument [
In all these uses, we can reason and argue about emotional content. For
example, to counter an emotional conclusion, one might question whether certain
actions, statements, or circumstances are consistent with a normative standard
under which the claimed emotion obtains. Where such inconsistencies arise, one
may counter-claim that the emotion did not normatively obtain, undermining
the claimants argument. Alternatively, there may be procedural moves, as in
where an emotional claim or emotional argument is ruled inadmissible in court.
In these various ways, we reason explicitly about arguments with emotional
content rather than simply ruling them out. As argued in [
One approach to modeling legal reasoning would be to model individual cases
or a corpus of legal cases, e.g. as in legal case-based reasoning [
We consider, in particular, California Criminal Jury Instruction CALCRIM No. 511 Excusable Homicide: Accident in the Heat of Passion, which establishes the conditions under which a homicide is excusable on the grounds of extreme emotion and cites cases for various points of the conditions, e.g. Substantial Emotional Distress De ned in People v. Ewing (1999) 76 Cal.App.4th 199, 210 [90 Cal.Rptr.2d 177].
To ground our analysis, we provide the relevant extracts from the two pages of the jury instructions for CALCRIM No. 511. We index clauses I - VII for reference, and we have omitted clauses irrelevant to our discussion relating to undue advantage, dangerous weapons, cruelty or unusualness of killing, intent to kill, great bodily injury, or criminal negligence: [I] CLAIM: The defendant is not guilty of (murder/ [or] manslaughter) if (he/she) killed someone by accident while acting in the heat of passion. Such a killing is excused, and therefore not unlawful, if, at the time of the killing: { 1. The defendant acted in the heat of passion; { 2. The defendant was (suddenly provoked by <insert name of decedent>/ [or] suddenly drawn into combat by <insert name of decedent>); { 3 - 7 indicate other, non-emotional conditions. [II] A person acts in the heat of passion when he or she is provoked into doing a rash act under the in uence of intense emotion that obscures his or her reasoning or judgment. The provocation must be su cient to have caused a person of average disposition to act rashly and without due deliberation, that is, from passion rather than from judgment. [III] Heat of passion does not require anger, rage, or any speci c emotion. It can be any violent or intense emotion that causes a person to act without due deliberation and re ection. [IV] In order for the killing to be excused on this basis, the defendant must have acted under the direct and immediate in uence of provocation as I have de ned it. While no speci c type of provocation is required, slight or remote provocation is not su cient. Su cient provocation may occur over a short or long period of time. [V] It is not enough that the defendant simply was provoked. The defendant is not allowed to set up (his/her) own standard of conduct. You must decide whether the defendant was provoked and whether the provocation was su cient. In deciding whether the provocation was su cient, consider whether a person of average disposition would have been provoked and how such a person would react in the same situation knowing the same facts. [VI] The People have the burden of proving beyond a reasonable doubt that the killing was not excused. If the People have not met this burden, you must nd the defendant not guilty of (murder/ [or] manslaughter).
The instructions also provide the duty of the trial court to give the instructions, related CALCRIM instructions, authorities (penal codes, case citations, secondary sources), and related issues (distinction between excusable, voluntary, and involuntary manslaughter).
[I1.] introduces the heat of passion element, which is clari ed (somewhat) in [II] and [III] as a violent or intense emotion that interferes with rationality. In [II] and [V], the provocation must be su cient to interfere in the rationality of a person of average disposition. [I2.] and [IV] highlight temporal dimensions: the provocation must be sudden (or combative) and be temporally close to the o ending action; while the temporal extent of the overall provocation is underspeci ed, presumably the nal \trigger" provocation is sudden. In [VI], the proof standard beyond a reasonable doubt is use to decide whether the killing was not excused; that is, if there is some reason that the killing was excused based on the conditions, then the jury should pass down this decision.
Having presented the elements of legal reasoning we model, we turn to outline computational models of emotions.
There has been substantial research on computational modeling of emotions in
agents and in modeling the concerns of others. The Ortony, Clore, and Collins
(OCC) model of the emotions [
[
For our purposes, it is not only necessary to represent the emotions of
individual agents, but also to be able to model the emotional representations of
others, particularly the defendant and the abstract person of average disposition
referred to in CALCRIM No. 511 since these are compared in giving a decision.
[
Argumentation schemes describe normative, presumptive, defeasible reasoning
patterns [
One example argument pattern is Argument from Distress.
Premise 2: Agent y's bringing A will relieve or help to relieve this distress. Conclusion: Agent y ought to bring about A.
There are various objections one might make about this argument: x is not in distress; even if y brings about A, it will not relieve this distress; it is not possible for y to bring about A; or, there are negative side e ects to bringing about A that preclude bringing it about. If one agrees with one or more of these objections, then the presumptive conclusion does not hold, and the argument is defeated. The objections might, in a dialogue, be cast as questions such as Is it the case that x is in distress?, where the negative answer introduces the objection, while the positive answer upholds the presumptive conclusion.
In this scheme, the emotional term distress appears among the premises;
that is, we do not have an argument for distress, where a statement such as
Agent x is in distress is the conclusion of an argument which follows from some
speci ed premises. While there are many other argumentation schemes that have
emotional terms among their premises, e.g. Threat, Fear Appeal, Danger, Need
for Help, and Distress [
In a legal setting, as outlined in section 2.2, determining whether the emotion normatively and plausibly holds or not is crucial to the legal decision. Not only must the premises be supported with reports and evidence from the defendant and witnesses, but also a COO must be constructed for that emotion that represents the person of average disposition. The emotional models for both the defendant and the COO for the person of average disposition are compared. It may, in addition, be argued that the defendant and COO models must be relative (e.g. child, psychologically abnormal, unusual circumstance, etc), subclassing the person of average disposition relative to the defendant's class. It is also worth noting as an aside that argumentation schemes with such emotional terms among their premises may also be considered rhetorical schemes which are used to persuade others. For example, Argument from Distress might be used as an argument by a prosecuting attorney that the jury ought to make some particular decision in a case. As part of this, the attorney would construct a COO model of the individual bearing the distress. Alternatively, in an Argument from Fear Appeal, the jury members' own concerns might be o ered as a reason for making a decision, thus requiring the prosecutor to model the jurors' hypothetical concerns.
Another important scheme in [
Premise 2: Each one of a set of accounts A1,..., An explains D.
Premise 3: Ai is the account that explains D most successfully.
Conclusion: Ai is the most plausible hypothesis in the case.
This is particularly useful in a legal setting where from known facts and several candidate theories, we reason to a plausible hypothesis, from which some legal decision will follow. Emotional conclusions may appear as parts of the accounts. For example, given as a fact that a perpetrator murdered a victim, the particular emotional context of the act may be signi cant in the legal judgment. If the best account for the murder includes a signi cant negative, shocking event which might (in the person of average disposition) induce emotional distress (even where this is not claimed by the defendant), this might be a mitigating factor in the judgment, deciding in favour of excusable homicide; alternatively, if no such abductive argument to an emotional state can be made, the absence of an emotion might be an aggravating factor. There are a range of objections one can raise for abductive arguments concerning the facts, the accounts for the facts, the success ranking, etc..
While argumentation schemes for emotions have not been discussed in the
literature, the role of emotions in the course of arguing has been. In [
[
As outlined in section 4, emotions in the context of argumentation have been
regarded as unargued for premises or as adjuncts in reasoning. However, as
claimed in [
We introduce emotional argumentation schemes, where emotion terms are the conclusions of argumentation schemes and follow from premises which are given by the OCC. Thus, as with other defeasible arguments, we can argue for or against emotional arguments. These emotional conclusions may then serve as premises of other arguments such as Argument from Distress or as components of such premises as in the abductive argumentation schemes.
As we do not have the space in this paper to give analyses of all possible emotional argumentation schemes, we provide one illustrative example which represents the elements taken from the full analysis of the emotion anger in the OCC. We have been concerned to represent the key clauses of CALCRIM No. 511, particularly: { Heat of passion. { Su cient provocation. { Sudden provocation. { Temporal proximity between provocation and o ending action. { Beyond reasonable doubt.
The objective of reasoning about these elements is to determine whether or not the defendant was irrational at the time of committing the o ending action. Our strategy has been to identify sub-arguments which form a tree of justi cation, linking conclusions of one argument with premises of another till we conclude with the root of the whole argument. In these schemes, the root conclusion is Agent x was irrational at the time of doing action a3, which is because Agent x was in the heat of a passion that interfered with rationality.
We relate the schemes here to the OCC in that we take into consideration
the concepts and relationships the OCC uses to explain emotions. The OCC has
formulae which calculate, from the values of several variables, the values of other
variables, e.g. intensity; in addition, there are complex issues about decay rates.
For our purposes, we do not provide a full analysis, including arguments and
formulae, for all these elements. In addition, the OCC and related work analyse
a spectrum of emotions in a range of degrees, while we are only interested in
creating arguments relevant to CALCRIM No. 511. In the following, premises
are introduced which would themselves require further argumentation and
eventual grounding in some base model of the emotions (for related treatments of
argumentation and semantic models see [
The schemes we introduce below would be used in several di erent ways: forward or backwards/abductive inference; comparing the emotional states and actions of the defendant to those of the person of average disposition. The comparison may give rise to further schemes and objections, which we do not introduce here.
As we are providing defeasible argumentation schemes, used in context where
knowledge is partial or perhaps inconsistent, there may be a variety of ways
to defeat the arguments: one may object directly that some premise (or the
conclusion) is false, giving the premises from which this objection presumptively
follows; one may object that while a premise is not false, it is insu ciently
supported, then providing an argument with the selected premise as conclusion,
but the argument itself has a falsi able premise; one may argue that the scheme
is inapplicable in a particular circumstance; one might cite exceptions which
hold, so the presumptive conclusion does not obviously follow. We leave implicit
these various ways of arguing against the schemes. However, these various ways
to attack the scheme represent the distinct ways that the arguments can be
attacked, moving closer to the goal of making such reasoning explicit and formal
[
Premise 1a: Agent y performs action a1.
Premise 1b: Action a1 highly con icts with the standards of Agent x. Conclusion c1: Agent x highly disapproves of Agent y's highly blameworthy action a1.
Premise 2a: Agent y performs action a2.
Premise 2b: Agent x intensely desires goal g.
Premise 2c: Action a2 results in not g.
Conclusion c2: Agent x is intensely displeased that not g holds.
Premise 3a: Agent x highly disapproves of Agent y's highly blameworthy action a1.
Premise 3b: Agent x is intensely displeased that not g holds. Premise 3c: The action a1 which Agent y performed is action a2 which results in not g.
Conclusion c3: Agent x was intensely angry at Agent y with respect to action a1.
Premise 4a: Agent x was intensely angry at Agent y with respect to action a1.
Premise 4b: Agent x performs action a3, which is not equal to action a1. Premise 4c: Action a1 happened in close temporal proximity to action a3. Premise 4d: Action a1 was sudden and highly unexpected by Agent x. Conclusion c4: Agent x was emotionally overwhelmed while doing action a3.
Premise 5a: Agent x was emotionally overwhelmed while doing action a3. Premise 5b: Being emotionally overwhelmed precludes being rational. Conclusion c5: Agent x was irrational at the time of doing action a3.
The schemes for Disapproval/Blameworthy and Intense Displeasure are used to argue for the conclusion of Intense Anger Scheme. The Emotionally Overwhelmed Scheme uses the intense anger conclusion along with temporal proximity and suddenness to conclude that the agent is emotionally overwhelmed. The Irrationality Scheme uses this conclusion along with a premise about the relationship between emotionality and rationality to conclude that the agent was irrational. This last conclusion is the target required (for our purposes) for excusable homicide { the killer was in the heat of passion, so not rationally in control of (or responsible for) his actions. Of course, a range of other conditions (not given) are required as well since the killing must also be accidental. Finally, for the burden of proof to be satis ed, there ought to be no reasonable means to defeat these arguments for irrational behaviour.
A fully spelled out range of argumentation schemes would be more extensive than these several schemes and include reasoning about the various elements of the OCC, the COO, the comparison between the defendant and a person of average disposition, auxiliary supporting evidence, and reported bodily states. Nonetheless, our analysis gives a clear indication of how emotional argumentation schemes can constructed, linked to further arguments, such as the relationship between emotionality and rationality, and elaborated further. In our view, a key advantage of presenting emotional argumentation schemes is not only the explicitness and clarity, but that we can introduce objections at key points which undermine the presumptive conclusions. Such objections are key in legal arguments and reaching judgments.
We propose to continue to research into the many facets of emotions in legal
reasoning so that they may be better understood and used in argumentation
schemes and argumentation frameworks. Achieving this would facilitate a
discussion of the relevant emotions present in the case by the judge, jury and
lawyers rather than dismissing them as ad hoc arguments. One potentially
useful approach is to use the argumentation schemes we have introduced along with
argumentation schemes used to argue about stories and criminal evidence [
We have here presumed the OCC and COO accounts of the structures of the emotions rather than providing them explicitly either as formulae or as argumentation schemes. It remains to be developed how to account for intensity, decay, and the role of moods which alter the parameters. Nor have we provided argumentation schemes for the spectrum of emotions. Similarly, our schemes may need to be enriched with other aspects of reasoning about the emotions that are relevant in a legal context. This said, argumentation schemes along the lines such as we have provided do seem plausible as representations of emotional arguments in legal settings; they also provide an extensible and exible structure for further development. 7
The second author was supported by the FP7-ICT-2009-4 Programme, IMPACT Project, Grant Agreement Number 247228. The views expressed are, however, those of the authors and should not be taken as representative of the project.