INFO282 Exam Review

A strategy, proposed and defended by Daniel Dennett, for understanding and predicting an entity's behavior.

the physical stance: To predict the behavior of a given entity according to the physical stance, we use information about its physical constitution in conjunction with information about the laws of physics. Suppose I am holding a piece of chalk in my hand and I predict that it will fall to the floor when I release it. This prediction relies on (i) the fact that the piece of chalk has mass and weight; and (ii) the law of gravity.

the design stance: assume that the entity in question has been designed in a certain way, and we predict that the entity will thus behave as designed. Clock alarm set to 8:30am, you will assume the clock alarm will then go off at 8:30am as usual.

the intentional stance: When making predictions from this stance, we interpret the behavior of the entity in question by treating it as a rational agent whose behavior is governed by intentional states. In contrast with the other two. STEPS: Treat object X as rational agent, Determine what beliefs X ought to have, given its place and purpose in the world, Determine X's desires, Finally on the assumption that X will act to satisfy some of its desires in light of its beliefs, predict what X will do.

• How to specify what does not change in a logical system when actions are applied. 
• Frame axioms sometimes present a serious problem called the frame problem. The challenge of representing the effects of action in logic without having to represent explicitly a large number of intuitively obvious non-effects. Using mathematical logic, how is it possible to write formulae that describe the effects of actions without having to write a large number of accompanying formulae that describe the obvious non-effects of those actions?

Is it possible, in principle, to limit the scope of the reasoning required to derive the consequences of an action?

For instance, an object’s color is unaffected by picking things up, opening a door, using the phone, making linguini, ect.

Reasoning about situations, and the actions that change them. very verbose.

• Situation do(a,s0)
• the situation resulting from performing action a in s0

• Situation do(a,do(a,s0))
• situation after performing a twice

a predicate which changes from situation to situation, i.e it has a situation-argument. predicates whose values vary based on situation. situation is always an argument, such as OnTable(box,s).

Rules of the form  <a:B/S> meaning S should be believed if a is believed and it is consistent to believe B. This leads to defeasible reasoning when the KB is updated with some fact inconsistent with B.

• Leads to successor state axioms (one per fluent, completely characterizes fluent in its successor state).
• If a fluent is not mentioned in an effect axiom for an action a, we would not know anything about it in the situation do(a,s). Necessary to know what fluents are unaffected by performing an action. Example Formula: dropping an object does not change its color, but it can change after painting.
• color(x,c,s) -> color(x,c,do(drop(r,x),x));
• These types of formulas limit or frame the effects of actions.

the two formulae above only license the conclusion that Position(A, Garden) holds. This is because they don't rule out the possibility that the colour of A gets changed by the Move action. The most obvious way to augment such a formalisation so that the right common sense conclusions fall out is to add a number of formulae that explicitly describe the non-effects of each action. These formulae are called frame axioms.

Poss(a,s) states whether or not a is possible to do in state s

• Precondition - which fluents / predicates must hold for an action to be performed.
• Effects - fluents changed as result of action

An axiom or postulate is a statement that is taken to be true, to serve as a premise or starting point for further reasoning and arguments.

We can formalize a completeness  assumption about the effect axioms we have for a fluent. Given 2 formulas, assume they characterize all the conditions under which an action a changes the value of fluent F.

Informally, these axioms add an "only if" component to the normal form effect axioms. F.eks you can rewrite an axiom which says that F is made true if setF (by set, I mean the mathematical symbol that is called "product" that looks similar to a pi symbol) holds. Using explanation closure, rewriting slightly to say, F is made true only if setF holds.

Precondition Axioms: one per action

Successor State Axioms: one per fluent

STRIPS planning, forward state-based search. Works by progressing the initial world model forward until we obtain a world model that satisfies the goal formula. As long as we visit each state only once we will always find a solution if it exists.

When a fact is only implicitly represented.  We say that the propositions represented by a set of sentences S entail the proposition represented by a sentence p when the truth of p is implicit in the truth of the sentences in S. In other words, if the world is such that every element of S comes out true, then p  does as well. All that we require to get some notion of entailment is a language with an account of what it means for a sentence to be true or false.

True

False

No. Reification is a way to make things real. In terms of logic it means to turn predicates into objects. Formally, we define a resolution derivation of a conclusion from a set of premises to be a finite sequence of clauses terminating in the conclusion in which each clause is either a premise or the result of applying the Resolution Principle to earlier members of the sequence. Think of the line chart to clear pos and neg to get the empty set.

Turns FOL sentances / formulae into Conjunctive Normal Form

An algorithm for testing the satisfiability of a set of propositional clauses. To check the validity of a first-order logic formula using a resolution-based decision procedure for propositional logic. CNF.

false, STRIPS allows one action at a time

When we happen to know that a polar bear has been rolling in the mud, or swimming in an algae ridden pool, or playing with paint cans, then we may not be willing to conclude anything about its color; but if all we know is that the individual is a polar bear, it seems perfectly reasonable to conclude that it is white.

Note, however, that just because we don’t know that the bear has been blackened by soot, for example, doesn’t mean that it hasn’t been. The conclusion does not have the guarantee of logical soundness; everything else we believe about polar bears could be true without this particular bear being white. It is only a reasonable default. That is to say, if we are pressed for some reason to come to some decision about its color, white is a reasonable choice. We would be prepared to retract that belief if appropriate evidence were encountered later. In general, this form of reasoning, which involves applying some general though not universal fact to a particular individual, is called default reasoning.

Default reasoning is a form of nonmonotonic reasoning where plausible conclusions are inferred based on general rules which may have exceptions (defaults).

AxEy(Man(x) & Woman(y) & Likes( x, y ) -> Likes(alex, x ))