Frame Problem in AI

Introduction

The frame problem originated as a narrowly defined technical problem in logic-based artificial intelligence (AI). Nonetheless, it was expanded upon and given a more nuanced interpretation by philosophers of mind. In the 1980s and 1990s, an engaging and occasionally tense debate arose from the conflict between the idea's inception in the labs of AI researchers and its treatment by philosophers. However, because the specific technical issue has been essentially resolved, current debates have tended to be less concerned with issues of interpretation and more concerned with the implications of the wider frame issue for cognitive research.

This article will start by examining the frame problem in its technical form so that the reader may better comprehend the problems. Then, it will be looked at how some philosophers have reinterpreted the issue. An evaluation of the frame problem's current significance will be made in the article's conclusion.

What is the Frame problem in AI?

The frame problem in artificial intelligence refers to a problem with employing first-order logic (FOL) to communicate facts about a robot in the real world. First-order logic (FOL) refers to logic in which the predicate of a sentence or statement can only refer to a single subject. It is also known as first-order predicate calculus or first-order functional calculus. Traditional FOL involves the use of numerous axioms that only imply that items in the environment do not change haphazardly in order to represent the state of a robot. In a FOL system, additional axioms are required to make inferences about the environment (for example, that a block cannot change position unless it is physically moved). The frame problem is the problem of finding adequate collections of axioms for a viable description of a robot environment.

This problem was defined in the 1969 paper Some Philosophical Issues from the Standpoint of Artificial Intelligence by John McCarthy and Patrick J. Hayes. The formal mathematical problem served as the foundation for more comprehensive considerations of the challenge of knowledge representation for artificial intelligence in this study and many that followed. Questions like how to implement logical default assumptions and what constitutes common sense to humans in a virtual world. Subsequently, the phrase took on a larger meaning in philosophy, where it is defined as the issue of having to revise one's ideas in light of new information. The implicit assumption in the logical context is that everything else (the frame) stays the same and actions are normally explained by what they alter.

Axioms are self-evident factors, that can be easily overlooked and forgotten. Let’s imagine a simple task like making coffee. In making coffee you need a lot of information about the coffee, the coffee jar, the machine, carrying a coffee mug, friction, and so on. Humans have learned these things by growing up in this world. They involve self-evident factors, such as, if I hold the coffee jar in my left hand, I cannot carry the mug with my left hand. Unless I have freakishly big hands. Or unusually small mugs.

Problems Related to Frame Problem in AI

The Frame problem can lead to several issues that can impact the effectiveness of an AI system. Some of these problems are:

• The Qualification Problem: John McCarthy introduced the Qualifying Problem. It implies that one can never be certain that a given rule will be effective. It also implies that the robot may not always be aware of which rules to break in a certain circumstance. Changes in the environment can "confuse" the robot since some rules will no longer apply and new rules will be required before the old ones do.
• The Representational Problem: The inability to produce accurate truths about the present environment is the Representational Problem. How would one program the concepts of up and down, for instance? They relate to one another and cannot be summarised in a single direction. Successor-state axioms are applied to somewhat resolve this issue. A
• The Inferential Problem: Difficulty with examining the methods by which the world is judged is the Inferential Problem. There are two kinds of purposes. The General Purpose is to inspect the entire world of things that are changeable. The Special Purpose is to only inspect actions that can modify over a small area of surroundings.
• The Ramification Problem: This problem explains how behavior might lead to changes in the surroundings. For instance, a brick has been picked up and moved to be placed on its side in a different area using a robotic arm.
• The Predictive Problem: The Predictive Problem deals with the benefits of predictions. That is, it is uncertain if a given prediction will cause a positive change in the environment. If the change will not be positive, "either the laws or description of the given situation must be imperfect

How to Solve the Frame Problem?

Several solutions have been proposed to overcome the Frame problem in AI. Some of these solutions are:

The Non-Deductive Approach

This strategy focuses on producing decisions that resemble human thought processes. Unfortunately, it has not yet been successful to programme human cognition.

The Deductive Approach

As opposed to the Non-Deductive Method, this method is free of psychological assumptions. All Knowledge can be expressed as axioms and use predicate calculus to arrive at conclusions where:

• S describes the task to be performed.
• A describes all actions in an environment. This is the frame part of the problem.
• Action A occurs and provides a conclusion for what A does in S.

This will only work in simple instances.

Problems:

• There is a specific action that is not specified in A, which results in an incorrect conclusion.
• If human behavior can be imitated, there is a claim that humans do not think in this way (comparison of A to S).

Frames & Scripts Approch (Minsky & Schank)

The Frames method of Minsky and Schank deals with categorizing and segmenting the world. Schank advises implementing the situations in scripts, but Minsky suggests doing so in frames. When a robot employs frames or scripts, it can develop routines for particular scenarios from which it infers the appropriate response.

Problems:

• It is hard to recover from a mistake.
• The robot could get stuck in a block that might not apply to all situations.
• It is uncertain how large each category should be.

Develop Experience Approach (Hume)

This strategy addresses the concept of "looking before you leap". The result produced by planning the next action based on an assumption could be used to broaden our understanding. We utilize the phrase "look before you leap" to create habits. The robot can gain experience and, in a sense, learn from its mistakes in this way.

Problem:

The one problem with this approach is how the habits can be represented in the knowledge base since a rule could branch into many different paths or lead into the same ones. A knowledge base is a form of artificial intelligence (AI) that aims to capture the knowledge of human experts to support decision-making.

Ad Hoc

The Ad Hoc certainty factor technique only adds probability into decision-making. Ad Hoc certainty factors can be used to predict success probabilities, which could help prevent poor choices.

Problem:

Certainty factors themselves are a problem because they are relative to each other and often derived from opinion. That is one person's view of 60% may equal another person's view of 70%.

Rethink the Semantic Level (Patrick Hayes)

The only suggestion made by this method is to make decisions regarding the types of information to examine on "histories and processes".

Android Epistemology (Clark Glymour)

Introduced by Clark Glymour, based on an idea by Daniel C. Dennett, android epistemology is the idea of combining philosophy and artificial intelligence.

Problems:

• Why should this be used for something that is obvious?
• It is still a mystery if a robot can think like a human. If it can't, how can it think philosophically?
• If a robot could think like a human, it could make wrong decisions as well.

Circumscription Approach (McCarthy)

Using the circumscription approach, appropriate conclusions are "jumped" to. Things that have already been established by another rule should not be searched for by the system. This method makes use of heuristics.

Problem:

The problem with this approach is the question of exactly when this method should be executed in the inspections.

Causal Connection Approach (Patrick Hayes)

This method proposes enjoying things in the world based on their shared characteristics. The characteristics of the other object must also be checked if one object is altered and causally related to another. This makes it possible to create exceptions and new connections.

Problems:

• What determines if something is causally connected?
• If an object changes, and it is causally connected to another, all properties of the connected one may require verification.

STRIPS (Fikes & Nilsson)

The STRIPS method makes use of both deductive and non-deductive methods. A robot will essentially check its surroundings by examining everything in a deductive manner, and then it will inspect the altered environment in a non-deductive manner, which is required for planning. This strategy is thought to be the best recognized one even if it doesn't work to solve the frame problem.