Digital Financial Reporting

This blog post pulls together a set of information which helps someone understand the moving pieces of digital financial reporting:

• Expert systems: Expert systems are computer programs that are built to mimic human behavior and knowledge.  A computer application that performs a task that would otherwise be performed by a human expert.
• Digital Financial Report Creation Vision: A digital financial report creation tool is an expert systetm.  This provides an example of the types of benefits offered by such an application.
• Framework and Theory: A framework and theory is necessary.
• Artificial Intelligence: Artificial intelligence is used to make an expert system work.  The term "artificial intelligence" tends to be overloaded and means different things to different people.  This is a good definition of artificial intelligence as I mean the term: 
  • Artificial intelligence is an area of computer science that deals with giving machines the ability to seem like they have human intelligence; the power of a machine to copy intelligent human But  behavior.
• Two approaches to constructing artificial intelligence: There are two approaches to building artificial intelligence:
  • #1-Know how to 'write down the recipe' meaning that those with the knowledge of some business domain articulate the information in machine-readable terms. In order to effectively express information in machine-readable terms, knowledge engineering skills are necessary.  This document, Knowlege Engineering Basics for Accounting Professionals, summarizes important information related to knowledge engineering.
  • #2-Let the machine grow the knowledge itself.  Another term for this is deep learning.
• Artificial intelligence to be 'the norm' in law firms in 2020: This article predicts that law firms will be using artificial intelligence by 2020 (only 5 years away).  It does not provide a lot of specific details.
• Artificial intelligence meets the C-suite: (McKinsey) "Many of the jobs that had once seemed the sole province of humans—including those of pathologists, petroleum geologists, and law clerks—are now being performed by computers."
• Digital Financial Report Information Quality Improving: I have been tracking the quality of XBRL-based digital financial reports for three years, information quality has improved significantly.  XBRL-US has created the Data Quality Committee to improve quality.
• Digital Financial Reporting Part of Larger Trend - Digital Business Reporting:  A digital financial report is a type of digital business report.  Another name for a digital business report is 'semantic spreadsheet'.

It would be great to synthesize all of this information into one clear, concise document.

Jeff Hawkins, in a 16 minute Ted Talk, How brain science will change computing, explains why a framework is necessary and why a theory is necessary.  He provides this explanation by taking about brains.  I can distill the important message pointed out by this video as it relates to digital financial reporting in this concise statement:

What is conspicuously lacking is a broad framework let alone a theory on how to think about digital financial reports.

Here are some of the important things that Jeff Hawkins points out his video:

• "We don't need more data, we need a good theory."
• "Things look complicated until you understand them."
• "We have an intuitive, strongly held, but incorrect assumption that has prevented us from seeing the answer. There is something that we believe that is obvious, but it is wrong."
• "Intelligence is defined by prediction."
• "We experience the world in a sequence of patterns." 
• "Real intelligence is a sequence of patterns."
• "Successful prediction is understanding."
• "Intelligence is making predictions about novel events."
• "Memory of high-dimensional patterns"
• "Memories are stored and recalled as a sequence of patterns"
• "Must be testable"
• "Must be buildable, if you cannot build it then you don't understand it."

For some reason, I intuitively understood that a theory and framework were necessary.  Rene van Egmond and I created a theory, Financial Report Semantics and Dynamics Theory, back in 2012.  That theory documents a framework.

The fundamental accounting concept relations are predictions.  I have been fiddling around with patterns that I have noticed in financial reports since 2000 or so.  I have used those patterns for many, many things; refining and tuning them along the way.  I have studied systems as part of the work toward my MBA way back in the 1980's and 1990's.

Seems like what Jeff Hawkins is saying is that the artificial intelligence community got it wrong.  (See minute 11 to about 13 in the video.

It seems to me like "memorizing" and "metadata" are the same thing.  Humans can memorize things, but these memories need to be put into machine-readable form for the computer to use.  So, these disclosures and these report frames seem like memories.  Using these memories helps you make predictions, "Because I see assets, because I see liabilities and equity, I predict that this is a balance sheet and not an income statement."

The biggest hurdle that I can see is summarized in this statement made by Jeff Hawkins:

We have an intuitive, strongly held, but incorrect assumption that has prevented us from seeing the answer. There is something that we believe that is obvious, but it is wrong.

[Another useful video, How Intelligent is Artificial Intelligence?]

Something that I have been trying to figure out is how good machine can be at figuring something out.  In trying to understand true capabilities I run up against terms such as 'artificial intelligence' and 'deep learning' and 'neural networks'.

Artificial intelligence I already discussed (see the link above).  Basically artificial intelligence is about getting a machine to mimic or simulate the behavior of a human.  For example, a calculator be it mechanical (hardware) or software is a machine that can mimic human behavior.

'Deep learning' and 'neural networks' seem to be related to ways of creating artificial intelligence.  Here are definitions of each of those terms:

• Deep learning: Wikipedia says: Deep learning (deep machine learning, or deep structured learning, or hierarchical learning, or sometimes DL) is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using model architectures, with complex structures or otherwise, composed of multiple non-linear transformations. (O'Reilly Media is publishing a book Fundamentals of Deep Learning)
• Neural network: A neural network is basically a computer system modeled after how the human brain and nervous system work.

If you look at the last sentence of the first section of the Wikipedia definition of 'deep learning' it says: "Alternatively, deep learning has been characterized as a buzzword, or a rebranding of neural networks."  Also, go look at Chapter 1 of the O'Reilly Media book, the chapter title is "The Neural Network".

This Basic Introduction to Neural Networks provides this explanation which sheds a lot of light on what neural networks are:

What Applications Should Neural Networks Be Used For?

Neural networks are universal approximators, and they work best if the system you are using them to model has a high tolerance to error. One would therefore not be advised to use a neural network to balance one's cheque book! However they work very well for:

•capturing associations or discovering regularities within a set of patterns;
•where the volume, number of variables or diversity of the data is very great;
•the relationships between variables are vaguely understood; or,
•the relationships are difficult to describe adequately with conventional approaches.

It seems to be the case that a 'neural network' is an architecture or approach to getting work done.  It is an alternative or compliment to currently used techniques.

People seem to have these delusions related to what you can do with a neural network.  Basically, there seems to be a lot of hype going around.  Slick marketing gimmicks lead you to believe that neural networks will do magical things for you without you having to put in any effort.  Their neural network is best.  But this is only hype.

Which type of network means far, far less than the model that you create and feed the neural network with.  That is the key, the metadata.

This SlideShare presentation, Deep neural networks, discusses the importance of unsupervised pre-training (slide 19) and supervised fine-tuning (slide 26) when constructing a system.

McKinsey published an article Artificial intelligence meets the C-suite.  That article makes the statement,

Many of the jobs that had once seemed the sole province of humans—including those of pathologists, petroleum geologists, and law clerks—are now being performed by computers.

(Don't forget this article about how artificial intelligence will impact law firms.)

Very interesting stuff.  Clearly there is a lot of hype and over-stating what a machine will be capable of doing.  But you cannot write off 100% of what people are purporting.  There is some truth buried in there.  Understanding how to make use of this technology is really important so you don't get fooled by those slick marketing gimmicks.

Stay tuned!

Artificial intelligence to be 'the norm' in law firms in 2020, is the title of an article claiming that all law firms will need to leverage artificial intelligence in the future as a standard technology investment. 

So, my first question is this: how will artificial intelligence impact professional accounting firms?  My second question is: what exactly do they mean by 'artificial intelligence'?

Merriam-Webster provides this definition of artificial intelligence:

: an area of computer science that deals with giving machines the ability to seem like they have human intelligence

: the power of a machine to copy intelligent human behavior

Basically, artificial intelligence is about making a machines mimic or simulate human intelligence.  There is no magic involved.  Machines, long before computers, have simulated human behavior.  Take the Babbage difference engine.  That was basically a mechanical calculator.  All hardware.  Computers are machines.  But what makes a computer unique is that while you do have some hardware, you also have software which makes changing what the machine does significantly easier.  That allows for the machine to change what it does because you can easily change the instructions.

Undoubtedly computers will have an impact on the work practices of accountants.  While I personally don't like the term artificial intelligence because the term is somewhat overloaded and it means different things to different people; I do believe computers will have a significant impact on financial reporting and the financial report.  I prefer to describe how this will happen as the creation of expert systems which leverage the structured nature of information which will enable new ways to create and work with financial reports.

The article goes on to give these steps to deploying the technology:

"Number one, you've got to get that artificial intelligence is going to fundamentally change the market. Number two, you've got to be able to deploy the technology - and that means you understand what's the right technology, what's available and how to use it. Number three, you've got to use it."

I completely agree that step 1 is "getting" that a fundamental change is about to take place.  And I agree that step 2 is "understanding" so that you will not be fooled by those trying to sell you a bill of goods.  And I agree with step 3 that "use it" is the best way to learn it and understand what the technology will and will not be capable of doing.

US public companies are getting the opportunity to have a look at the future of financial reporting.  In fact, they are blazing the trail.  Most don't even understand that that is what they are doing.  They see the SEC mandate to report using XBRL as more of an expensive and useless nuisance than experimentation with a cutting-edge technology.  But that will change; in fact, it is changing already although slowly.

If you want to better understand how law firms, professional accounting firms, and financial reporting will likely change; the document Knowlege Engineering Basics for Accounting Professionals is helpful.

The American Institute of Certified Public Accountants (AICPA) took bold leap back in 1998 when it started its journey toward digital financial reporting.  Perhaps lawyers and other professionals can learn from what accountants have been able to put together with the help of some information technology professionals.  While digital financial reporting still has some rough edges, a lot has been learned.

Digital financial reporting is not only inevitable, it is imminent.  Could it be the norm by 2020?

The FASB released a proposed 2016 US GAAP Financial Reporting Taxonomy in XBRL and is taking comments until October 31, 2015.  Taxonomies such as the US GAAP Financial Reporting XBRL Taxonomy are important tools in our digital age and as I mentioned in the document Knowledge Engineering Basics for Accounting Professionals, taxonomies overcome the four major obstacles of getting a computer system to perform work:

• Business professional idiosyncrasies: different business professionals use different terminologies to refer to exactly the same thing.
• Information technology idiosyncrasies:  information technology professionals use different technology options, techniques, and formats  to encode information and store exactly the same information.
• Inconsistent domain understanding of and technology's limitations in expressing interconnections: information is not just a long list of facts, but rather these facts are logically interconnected and generally used within sets which can be dynamic and used one way by one business professional and some other way by another business professional or by the same business professional at some different point in time.  These relations are many times more detailed and complex than the typical computer database can handle. Business professionals sometimes do not understand that certain relations exist.
• Computers are dumb beasts:  Computers don't understand themselves, the programs they run, or the information that they work with.  Computers are dumb beasts. What computers do can sometimes seem magical.  But in reality, computers are only as smart as the metadata they are given to work with, the programs that humans create, and the data that exists in databases that the computers work with.

The US GAAP Financial Reporting XBRL Taxonomy is a formal specifications.  Formal specifications are precise, concise and unambiguous.  Formal specifications are communications tools.  Because taxonomies use machine-checkable notation, a wide variety of automated checks can be applied. The disciplined approach of using formal specifications means that subtle errors and oversights will be detected and corrected.

An example of these checks is the fundamental accounting concept relations that I test each month.

Taxonomies both describe the information being worked with and verify information consistency against that description to avoid information quality problems or inconsistencies.  When creating information it is important to verify that what has been created is consistent with the expected description.  When consuming information it is important to understand that the information being consumed is consistent with the expected description. Remember: nonsense-in-nonsense-out.

The first two obstacles which related to the problem of business professional idiosyncrasy and technical idiosyncrasy are overcome by using a taxonomy to standardize terminology.  Rather than using arbitrary  terminology to express information about some business domain, standard terminology is used.  This includes selecting the appropriate important terms and defining the terms in a deliberate, rigorous, clear, logically coherent, consistent, and unambiguous manner.  Care is taken to precisely and accurately reflect reality using standard terms.

The third obstacle of expressing the rich logical interconnectedness of facts within and across business systems can be overcome by using general ontological theories disciplined, methodical, and rigorous approach to structuring the relations between terms.  Meaning, expressed using machine-readable taxonomies, must be exchangeable between business systems not just used within your one system.
Beginning with a rigorous and logically coherent specification of the theoretical information to be implemented makes it possible to address the problems of human idiosyncrasy.

Given the idiosyncratic tendencies of business professionals; interpretations which reflect the arbitrary peculiarities of individuals can sometimes slip in or mistakes can be made when expressing such terminology.  Further, parts of our understanding of a business domain can be incorrect and even evolve, improve, or simply change over time.

If different groups of business professional use different terminology for the same concepts and ideas to express the exact same truths about a business domain; those business professionals should be able to inquire as to why these arbitrary terms are used, identify the specific reasoning for this, and specifically identify concepts and ideas which are exactly the same as other concepts and ideas but use different terminology or labels to describe what is in fact exactly the same thing.  But to also understand the subtleties and nuances of concepts and ideas which are truly different from other concepts and ideas.

If idiosyncrasies result only in different terms and labels which are used to express the exact same concepts and ideas; then mappings can be created to point out these different terms used to express those same concepts and ideas.  Such mappings make dialogue more intelligible and could get groups to accept a single standardized term or set of terminology for the purpose of interacting with common repositories of business information, such as the SEC's EDGAR repository of public company financial reports.

If the difference in terminology and expression are rooted in true and real theoretical differences between business professionals, and the different terms express and point out real and important subtleties and nuances between what seemed to be the same terms; then these differences can be made conscious, explicit, clear, and therefore they can discussed, in a rigorous and deliberate fashion because the differences are consciously recognized.

The best indicator of whether a taxonomy is working or not is whether it achieves the goal; whether the comunication between the creators of information and the consumers of information is successful.

Consider these ideas and consider commenting on the proposed US GAAP Financial Reporting XBRL Taxonomy.