%\newtheorem{theorem}{Theorem} %\newtheorem{corollary}[theorem]{Corollary} %\newtheorem{definition}{Definition} \chapter{Promises, obligations, commands and requests}\label{appa} Throughout this book, we use the term `promise' with a more or less constant intent. Nevertheless it turns out to be difficult to provide a clear definition without making it unreadable and sacrificing intuition. Thus, in this Appendix, we would like to expand on the manner by which we arrive at our definition \ref{def_micropromise} of chapter \ref{chap_notations} n commitand its variations by extending the philosophy of chapter \ref{start}. \section{Ontology} We proceed from the following: \begin{definition}[Promise] \label{def1} A public announcement (made by an agent called promiser and performed as an act of free will) of an intention, of a plan to be executed or of the expectation of a fact to hold true either in the past the present or the future (the promise body usually composed of a type or quality and a size or quantity). Among the audience (scope) of the promise there may or may not be one or more specifically designated agents who are the intended beneficiaries of the fact or action implied by the body. It is conventional for the announcement being classified as a promise is that it should effect expectations within the promisees which are positively valued as such. \end{definition} For use in engineering, computing and other technical disciplines, where promisers and promisees are likely to be inanimate agents, the following weakened forms of this definition will be relevant: \begin{enumerate} \item A promise may not have promisees, in which case it is merely an announcement of a plan or of an expected or existing state of affairs to the agents in its scope. A promise to a promisee has in addition to being an announcement of a plan or state of affairs the aspect that the promiser is aware of the promisees interest in the promise being kept, as well as the fact that the promisee may expect that this awareness is seriously taken into account when the promiser decides not to keep the promise. \item A promisee's changed expectations, on receiving a promise, need not be made explicit and the change need not be made in a positive direction. A threat is a promise made with generated expectations valued negatively rather than positively. Thus threats are included in promises, an conversely for that matter. \item There may be different promisees, and these may value their change in expectations differently. What is an attractive promise to some may be a threat to someone else. \item Rather than speaking of an act of free will, both issuing the promise and acting in compliance with its body should be properly classified as {\em autonomous behaviour} of the promising agent. \end{enumerate} Below we will provide two weaker definitions of promises (that is, by qualifying more kinds of announcements as promises), and definition \ref{def3} will be the final one for our purposes. \subsection{Notions related to promises} Lots of notions can be found that relate to the notion of a promise. Here is a listing. \begin{itemize} \item Obligation \item Promise \item Forced promise \item Autonomous promise \item Command (instruction) \item Request (an asking) \item Suggestion to perform a plan \item Announcement of plan \item Announcement of intention \item Plan (intention of plan, intention to execute plan) \item Intention \item Execution of plan (activity) \item Forced activity \item Compulsive activity \end{itemize} Besides these concepts that all have to do with speaking about plans and their execution either by the agents themselves or about other agents there are surrounding notions that come into play. \begin{itemize} \item expectation \item hope \item fear \item ambition \item conventional pattern of action \item will (free will) \item trust \item distrust \item positively valued expectation \item negatively valued expectation \item happy state of an agent \item partially happy state of an agent (positively valued aspect of state) \item unhappy state of an agent \item partially unhappy state of an agent (negatively valued aspect of state) \item agent life-cycle with at least the stages: birth, active phase, and death \item force exercised upon an agent \item objective \end{itemize} All of these notions may play a role when describing a promise. That makes the concept particularly complex. Our definitions of promise take into account only a limited set of aspects. \subsection{Why make promises} When working towards a definition of promises some examples are needed where the act of promising is convincing form of a rational behavior. Interestingly one can find examples where promise making cannot be replaced by other actions. Here is an example. \subsubsection{Posthumous publishing} Suppose that agent $B$ is approaching the end of his life and $B$ is aware of that fact. $B$ asks agent $A$ to take care of publishing his nearly finished book via publisher $E$ where all that needs to be done before submitting is one round of proofreading.\footnote{A classical related example is Vergil, who asked to be promised that his manuscript Aeneid be destroyed.} We assume that $B$ has no life after death, all goodies relevant to $B$ happiness must materialize during his life. The following scenario is now plausible: \begin{itemize} \item We assume that $B$ has no means available to arrange via his will that someone takes care of publishing the book, and that $B$ has no access to other authorities who may force $A$ to publish the book after $B$'s death by means of an immutable obligation for $A$ to do so. \item In response to $B$'s request, $A$ promises to $B$ to proofread the manuscript, to repair obvious mistakes and to submit it to publisher $E$ subsequently. \item As a consequence of hearing $A$'s promise $B$ reaches a partially ``happy'' state (regarding the expected fate of his most recent manuscript) because of the positive expectation that the promise will in fact be effected. \item The `happy' state of $B$ is a positive value even if it will not lead to any action observable by other agents. In other words it is to be valued positive as such that $B$ ends his life while being in a partially `happy' state. (The more ``happy'' the better so to speak.) \item $B$ has no possibility to check or test whether or not $A$ will keep its promise, \item For $B$ there is no other way to obtain this particular state of happiness before the end of his life than to ask for that promise to be made by someone who he can communicate with and in who he has sufficient trust. \end{itemize} Modifications can be easily imagined. In this case, however, the promise made by $A$ seems not to produce any good over and above what would have been achieved if $A$ made an announcement of plan with $B$ in the scope of that announcement. \begin{itemize} \item In a modified example where $B$ is still enabled to edit his will $B$ may try to force $A$ into making the promise by making the execution of part of his will towards $A$ dependent on that. \item In yet different circumstances $B$ may try to create an obligation for $A$ to execute the promise body by asking some higher authority to issue a command towards $A$ for doing so. \item Assuming that after his death $B$ is insensitive to any further event, (a consequence of the assumed absence of life for $B$ after his death) the thereafter merely $A$'s happiness is at stake if he fails to live up to his promise. If more agents were in scope of the promise $A$ may have additional incentive to live up to it. \item This example provides an example of a promise where the objectives of $B$ cannot be reached by anything else than a promise being made. $A$'s mere announcement of plan will generate the same expectation inside $B$ and for that reason it qualifies as a promise in the sense of definition \ref{def1}. None of the other related notions appropriately describe what is going on. This example may be considered a proof of existence of the concept of promise as defined in \ref{def1}. \end{itemize} \subsubsection{Feedings a cat during its owner's vacation} Agent $B$ will be on vacation for one month and needs his plants watered. Unless this is properly done the plants won't survive. $B$ asks $A$ to do so and $A$ makes the promise to do so with only $B$ it scope of the promise. Now like in the previous example $A$ cannot force $B$ to live up to the promise during the period of his vacation. But there are significant differences: (i) after returning $B$ can see whether $A$ has done his job, (ii) if $A$ has not lived up to expectation $B$ may try to punish him for not keeping his promise (it is quite difficult to formulate in abstract terms that $B$ does neither have the power nor the intention to do so), (iii) after his trip $B$ can reward $A$ for keeping his promise (and by making the promise $A$ may develop the positively valued expectation that this will take place), (iv) $B$ might return sooner and see what is going on (if $A$ is not keeping his promise), (v) $B$ might lose his trust in $A$ while being on vacation and for that reason ask another agent to see into the matter. There seems to be no doubt that all these differences do not change the situation to such an extent that $B$'s announcement of plan fails to qualify as a promise. The reason for that is that under default circumstances $B$ is dependent on $A$ keeping is promise as an act of free will. Only if additional changes in the situation emerge the setting is changed in such a way that what started its life as a promise made autonomously as an act of free will turns into the forced execution of a plan against one's free will. Here we see that when checking that executing the promise body should be an act of free will a default logic may come into play. \section{Derived constraints on promises} \begin{enumerate} \item The absence of both obligation and force is essential for classification of an announcement of plan as a promise. Thus either the notion of free will is take as a primitive one or it is reduced to the absence of force and obligation. Free will being a complex concept itself its seems compelling to view both force and obligations as simpler and to replace in the definition of promise the free will requirement by a requirement that no force or obligation are relevant for the announcement or the execution of the plan in its body. This would be \begin{definition}[Promise] \label{def2}A promise is defined as in definition \ref{def1}, but now with the free will requirement replaced by the requirement that force or obligation are not involved. \end{definition} \item If $A$ is forced to execute plan P he cannot promise to do so at the same time (unless P is unaware of the force applied). The same holds for making the promise. Here we arrive at an unclear area: suppose $A$ is obliged to perform plan P but is unaware of that fact. The $A$ can be asked to promise to do so. Has $A$ made a promise. According to both definitions he is not. But he cannot see this fact for himself. Should we take into account the possibility that $A$ cannot reliably judge whether he is making a promise when he subjectively thinks he is doing so. This is less attractive and therefore we are led towards a further sharpening of our definition: \begin{definition}[Promise] \label{def3}A promise is as defined in definition \ref{def1} but now with the free will requirement replaced by the requirement that to the best of $A$'s knowledge force or obligation are not involved. \end{definition} This definition leaves open the possibility that $B$ asks $A$ to promise P while $B$ knows that he can force $A$ to do so and $B$ is unaware of that fact. Still, according to definition \ref{def3} the announcement qualifies as a promise for $B$. This subjective aspect reinforces the reduction of 'free will' to the absence of force and obligation. That can be easily weakened to the perceived absence of both. Instead the distinction between free will and perceived free will is harder to grasp. \end{enumerate} \section{Force, command and obligation} A request is complementary to an announcement of plan. $A$ request by $B$ to $A$ can be issued and it is reasonable that $B$ provides either an announcement of plan or a promise in its return. For an announcement of plan it is no requirement that force or obligation are absent or that any form of free will is established or assumed by either by $A$ or by $B$ or any other agent involved. An obligation differs from a request in the sense that the obligation for $B$ to perform plan P should be derived from a general rule that agents which match properties X should perform in accordance which Y and that for $B$ acting in accordance with Y implies to execute plan P. Obligations (for $A$) can exist without having been announced by any agent. A typical rule for creating obligations may be that $A$ should perform in compliance with the commands issued by $B$ (because $A$ is working under $B$'s supervision). This turns all commands issued by $B$ into obligations of $A$. At the same time it makes implausible the qualification as a promise of any announcement of plan made by $A$ towards $B$, insofar as these plans are with in the scope of $B$'s authority over $A$. Force is an other matter altogether. Force is not derived from a general rule but it comes into effect in specific circumstances. It may be questioned whether or not a human agent can ever be forced into any action at all. In the setting of engineering, e.g. computing, force is more easily imagined. An agent may be forced if some of its actions are given a very high priority in some context. But will assume the existence of force in the more general case. Coercion, a kind of force, is in place if an agent is made aware of negative consequences unless some plan is performed. These negative consequences lead to a state of unhappiness which is valued negatively. Indeed the example of a promise given above can be extended with a historic context in which $B$'s death is expected because he has not complied (which may very well have been an act of free will) with some command (for instance by A) and has been awarded a capital punishment for that reason. Still $B$'s unhappiness about the state of affairs can be diminished by the promise made by $A$ concerning finalization of $B$'s book. The qualification of $B$'s announcement to see to its publication as a promise is unaffected by the context in which an attempt to apply force led $B$ into the problems that provide an incentive to ask for A's promise. \section{Another promise} Suppose at the end of $B$'s life, $A$ makes $B$ the promise that $A$ will have a further life in another world. This generates a positive expectation inside A's mind and he ends his life in a `happy' state for that reason. Assuming that one does not believe in the second life in another world one may still acknowledge the state of happiness thus created for $B$. Are we to conclude that: (i) the announcement of future fact made by $A$ qualifies as a promise, (ii) because of the positive valuation by $B$ of the announced fact that generates a state of happiness which is for that reason to be valued positive as well the very act of making the promise can be evaluated positively as well. This is a difficult matter. $A$ promise is qualified as a deception if the promiser does not expect either the announced fact to hold true or the announced plan to be performed (depending on the form of the promise). As it stands the act of announcing a deceptive plan or fact is not to be valued positively as such in spite of the positive value assigned to it by the promisee. The matter is hard to resolve: if announcing eternal life in another world is the only possible way to increase $B$'s happiness at some stage, should $A$ for that reason do so? Perhaps some people adhere to a specific faith only for the reason that when this case arrives they can make the promise without hesitation. These decisions are an ethical matter. Nevertheless it seems defensible to exclude from qualification as promises those announcements of plan or fact for which no active (living) active agent, either now or in the future can ever assert whether the announced fact has come about or whether the announce plan has been performed. The the announcement of eternal life is not qualified as a promise but should be termed differently, e.g. as a 'religious promise', which fails to qualify as a promise just like a cancelled flight qualifies as a flight. In this way deceptions still qualify as promises in as far as the coming about of the announced fact or plan can be determined objectively in some time to come. This in turn can be taken as incorporated in the notions of plan and fact rather than in the definition of a promise as such. Thus the event of $B$'s eternal life in a second world would not be regarded an admissible 'fact' (but say merely a religious fact instead), thus allowing not to classify $A$'s announcement of $B$'s life after death as a promise. Again one might criticize this distinction because it is unknown now which forms of knowledge future generations will have available to them. Perhaps some day the death will appear to the living so that these must acknowledge their preservation after death. But now one may use default reasoning to argue that sources of knowledge are absent unless known to be present. A conceivable argument, e.g. that future generations will develop a 'religious radar' which enables them to spot the deceased in heaven cannot be refuted by any means, but default reasoning indicates that it should not be taken into account unless it has actually materialized. Such an argument cannot be used to qualify a supposed state of affairs in a promise body as a potential fact which is necessary for the qualification of the announcement as a promise. \section{Conclusion} We have shown to our satisfaction that promises exist. However, to arrive at an applicable definition, the concept must depend on that of force and obligation, in the sense that the absence of these must be required. We conclude for that reason that in a detailed analysis of the general concept of a promise, it cannot precede the introduction of any notion of force and obligation. However, a drastic simplification is achieved if the concept of autonomous action is used as a primitive instead. Now the absence of force or obligation is implicit in the notion of {\em autonomy}. It seems reasonable to use the following definition of a promise in the case of computing. \begin{definition}[Promise] \label{def4} A promise is as defined in definition \ref{def1} but now with the free will requirement replaced by the requirement that act acts autonomously both when making the announcement and when performing the plan implied by its body if any. \end{definition} When using this definition in engineering, e.g. computing, care should be taken that when notions of agent involved become more refined, at some stage, autonomy ceases to be a primitive notion and the more restrictive view such as in definition \ref{def3} should be preferred. Under this assumption its seems valid to consider a promise according to definition \ref{def4}, as a concept that is independent of the concepts of force and obligation.