Harekrishna,

I think that as stated, the answer to your question has to be "no, but...". There's an element of creativity in choice of strategy that an expert system can't replicate. However, in implementing that strategy, then you might be able to get a decent bit of help from an expert system, as long as the strategy itself is well defined enough. Strategy implementation is what your guided missile example is about. Some human decision-makers decide what to aim the missiles at (at least I hope they do!), and the missile then chases the target meeting the set criteria. Expert systems also are not much use for simulation, though you can use an ES to represent a human decision-maker in a simulation model using discrete-event simulation software like Witness or SIMUL8; I've been involved in research doing some of that.

Best wishes,

John E

The answer could be both yes or no.

As a vetrianarian and having had experience in designing expert systems for diagnosis of animal diseases, an expert system could certainly help narrowing diagnosis and even treatment. With the current explosion in human and animal health care information it is very difficult to keep up with all advances in medicine.

As regards business decision scenarios, ultimately the experience of automated stock market participation does indicate that when things go wrong, they really do destroy businesses. The problem is not with the technology but with the human capacity to build greater accuracy and precision in business decisions which has really too many unknown variables including human behaviour.

Also, as the recent episode on Qatar Airways new Airbus 350 indicates, expert systems can make faulty decisions. In the Airbus case though there was no harm caused, the dangers of a computerised, complex fly by wire system dependent on knowledge based systems to fly that can suddenly stop or do something unintended do become apparent.

Harekrishna,

I think that as stated, the answer to your question has to be "no, but...". There's an element of creativity in choice of strategy that an expert system can't replicate. However, in implementing that strategy, then you might be able to get a decent bit of help from an expert system, as long as the strategy itself is well defined enough. Strategy implementation is what your guided missile example is about. Some human decision-makers decide what to aim the missiles at (at least I hope they do!), and the missile then chases the target meeting the set criteria. Expert systems also are not much use for simulation, though you can use an ES to represent a human decision-maker in a simulation model using discrete-event simulation software like Witness or SIMUL8; I've been involved in research doing some of that.

Best wishes,

John E

Many general ways of solving the problem are known. After training algorithms in specific domain, we can think of replacing CEOs/decision makers by trained programs that can handle most routine situations.

Domain specificity is very important.

Dear Colleagues,

Good Day,

Guided Missile system is completely different, error corrections are done every (1/16 second) at least. That is done within complete feedback system.

So you can not map this to to the business decision scenario. unless you have to scale up the time and you know all the inputs and the outputs plus all the internal and the external disturbances with readiness for any losses (money spending for corrections!!!!).

PS. this is, merely, from a control engineering point view, and reflected to business.

Game theory is "the study of mathematical models of conflict and cooperation between intelligent rational decision-makers."(see link)

https://en.wikipedia.org/wiki/Game_theory

Expert systems are designed by human thinking, analysis and interpretations. As far as human brain and thinking process are concerned, in certain narrow domains of engg. and natural sciences they may consider all the parameters and constraints governing a decision making process. However, it is almost impossible to consider entire gamut of influencing parameters by human minds. That is where the expert systems fail.

it is possible guidelines leaves space to strategic decision makers to use empirical and heuristic to lower levels, since even complex management problems can be successfully.

Best regard. I am not an expert type guided missile systems. However, if the issue is about making decisions, I believe that is the field of strategy. Making decisions is an art that is not only based on rational aspects, there are intuitive aspects that influence. The way the elements of reality is categorized influence cognitive system that finally finishes building the map that the brain needs to make the decision. This is affected by the principle of bounded rationality, asymmetric information and uncertainty levels. Take decisions only mediated by computational logic does not guarantee hit the target.

My answer is Yes because we all know that computers mind is much faster then a human .if we give them better information of our current business strategies then they can give us better solution.probably heuristic search. 

The missile scenario and the CEO decision making business scenario are completely different problems, in my humble opinion, in terms of the type of data/information that are based on. The missile gets lots of quantitative/numerical data from sensors in real time (milliseconds?) and the problem is more an automatic control problem not a human-like decision making problem. In the business scenario you need to deal with complex knowledge of different sorts including common sense knowledge, knowledge based on lots of experience, etc. Therefore my answer is that they cannot be mapped.

Harekrishna,

A system that can help a decision maker is called a DSS (decision support system). The hypothetical expert system that you consider in your question  would be a (computerized) decision maker rather a DSS because the purpose of an expert system is to reproduce the reasoning capabilities used by a humain in solving a particular problem (here finding the or an "optimal" solution in the stategic decision-making problem of interest). Acquiring and representing the appropriate knowledge is still a major difficulty; should it be from only one expert, can it incorporate coherently the expertise of several experts, how to deal with the ineffability of tacit knowledge used in strategic decision problems ?

Besides there is still the issue of responsibility bearing in case of bad decision generated by the expert system. Who should be blamed in such a case:  the expert(s) who contributed to the knowledge base of the expert system, the knowledge engineer that coded the knowledge base, the user of the expert system?

The short answer is no. Even though a missile may look like its making sophisticated decisions these decisions are planned (by programming teams) based on complex statistical models (eg Kalman filters operating on the sensor- and dynamic data of the missile). They may even employ multi-scenario(viewpoints) models so that they can recover from countermeasures and the like. But their domain is their dynamics, their sensors and their objectives. (Some aspects of AI may be found in weapon systems but these are from ideas that can be found 1980/90s to cope with strategy and target aquisition -- tools such as KEE, knowledge craft , ART, Symbolics Lisp Machines etc).

An Ideal AI machine is one that could do its own planning and even that would be restricted to its own domain. Thus human decision support needs to span the human business domain. Attempts  to do this  in a lingusitic manner (eg rules, semantic networks etc)  usually fail because the domain of text, words language etc requires an underlying basis of sensor-derived experience to check the validity of inference. A purely textual basis is somewhat elusive! (I believe its impossible).

For the domain of Marginal farmers an analysis of that domain may yield sufficient parameters to build an "expert system" focussed on their needs. This will be  specific to their problem and  could include heuristics derived from modelling.

Years of AI tools sets (1980s primarily) proved one thing: generic solutions to problems can only be implemented after very specific modelling. After the modelling you don't really need the expensive AI toolset  - a simple computer language or even a spreadsheet could be sufficient.

Dear Harekrishna,

I think in case of replicating systems it's worth to try to build a model like the weighted graph or singleton in order to add some kind of heuristics or cross-product it with another probabilistic scheme, such scheme could actually build an additional finite state machine for simulation plan (with additional costs and/or events added), please refer to or read Dr. Buchi about it. Also in FSM theory I met cases, i.e. there are, when finite state graphs are built on the fly. In any case, this could be done relatively to the full-state graph or full-event experiment.

Additionally the heuristics could be applied on the fly, like ant colony optimization by M. Dorigo - you can read my first paper on this case, which popularizes this method.

I recommend to use regression method measurement scheme according to Fischer's methodology.

Kind Regards,

Mirzakhmet Syzdykov

Harekrishna,

I would first ask if the system is an expert system or a decision support system. Is the system working using rules in forward chaining (expert system) or using function minimization rules(decision support)?

To answer your questions both may be useful under different scenarios and different types of users. From what little detail I have I would consider your system a decision support for production optimization. An expert system would be used for decision making where the inference rules need to be traced through the chain of reasoning.

Hope this helps

Hello Harekrishna,

Here is my contribution to your question(s):

"We come across guided missile systems hitting the targets with guided paths and re-routing it as per the dynamic and real time situations."

++My experience has been that expert systems that support the military decision making map better to medical situations than to business situations. A disease is like an invading army that also tries to evade detection and destruction as targets. The similarities between military situations and business are less evident, although you should be able to reuse the upper ontology.

"Can this situation be mapped to the business decision scenario and create tools for the CEOs and decision makers to simulate and find real time solutions?"

++Like most of our colleagues have indicated, the short answer is no. This is mainly because tactical military situations require the kind of "decisive" responses with very short deadlines that leave little or no time for deliberation and analysis, whereas business decisions more frequently require research, analysis, and strategic thinking. The military and business mind sets are really at opposite ends of the tactical-strategic spectrum. 

"Are their such tools available meeting the requirements?"

++Again, the short answer is probably not. As I have indicated above, however, you may be able to reuse some components of an ontology, particularly the upper part, if you want to construct a decision-support system that is based on an expert system. You will likely need to re-engineer so much of the software and lower ontology that starting from a different point will be more productive. If you start with a business-oriented ontology as the basis for your knowledge base, you can save a great deal of time and money.

++As I have implied, expert systems and decision-support systems are not mutually exclusive. However, most expert systems that I have seen were designed to support decision making. How far they go in actually achieving this goal depends on their quality, ease of use, and how much experts use them. Expert systems used as DSS are not really designed to take the place of experts, but rather, to provide a valuable component in the information that experts use to make decisions.

"Could these be bench marked for replications cross various sectors including dairy dealing with marginal farmers?"

++The key here is "various sectors." The main sector that is most similar to military, as I have indicated above, is the medical sector. However, dairy farming is far afield from both the military and the medical sectors. Farming is better approached with a different mind set, and a lower ontology to match.

In summary, if you want a plow, beating your sword into a "plowshear" is an inefficient way to produce one. You would be better served to design the plow for the purpose of plowing, rather than to retrofit a sword. 

An guided missile system (GMS) operates on input information that is pre-determined, well-structured, quantitative. It has only one function or one kind of problem to deal with (target and destroy). The system can only cope with situations that it has been pre-programmed to cope with. Such situations have largely been mapped out and therefore a GMS works as well as it does although yet imperfectly.

Contrast this situation with the kinds of circumstances and problems a multi-tasking CEO must deal with and ask yourself the following questions. Does (s)he always deal with pre-determined,  pre-defined, well-structured information in all decisions (s)he makes? Does (s)he deal with a single problem situation at all times? Are decision problems (s)he deals with always foreseen? Is the information needed in each problem situation identical to other problem situations. Finally, does a CEO always rely on complete, accurate (within tolerances), quantitative input?

Asking such questions, the answers to many of which are "no" quickly leads to the realization that a CEO cannot be replaced by a machine. Finally, one also needs to ponder whether a GMS is a strategic decision maker or an operational and tactical decision maker. I would think it is not although, as a weapon, it may be used in a strategic manner by those who use them in dealing with others. But its own functionality can hardly be termed "strategic."

An guided missile system (GMS) operates on input information that is pre-determined, well-structured, quantitative. It has only one function or one kind of problem to deal with (target and destroy). The system can only cope with situations that it has been pre-programmed to cope with. Such situations have largely been mapped out and therefore a GMS works as well as it does although, imperfectly.

Contrast this situation with the kinds of circumstances and problems a multi-tasking CEO must deal with and ask yourself the following questions. Does (s)he always deal with pre-determined,  pre-defined, well-structured information in all decisions (s)he makes? Does (s)he deal with a single problem situation at all times? Are decision problems (s)he deals with always foreseen? Is the information needed in each problem situation identical to other problem situations. Finally, does a CEO always rely on complete, accurate (within tolerances), quantitative input?

Asking such questions, the answers to many of which are "no" quickly leads to the realization that a CEO cannot be replaced by a machine. Finally, one also needs to ponder whether a GMS is a strategic decision maker or an operational and tactical decision maker. I would think it is not although, as a weapon, it may be used in a strategic manner by those who use them in dealing with others. But its own functionality can hardly be termed "strategic."

I was asked some years if it would be possible to develop the Silent Talker system (Neural Network based)  to automatically identify and shoot a terrorist approaching a checkpoint. My response was that AI systems (apart from trivial applications) are not 100% accurate giving ride to both ethical and practical reasons for not doing so. This has in recent years been codified by the EU into the "man-in-the-loop" ethical requirement.