(Click on the presenters or scroll down to see presentation abstracts and author biographies)
- Welcome to DecisionCAMP-2022 by Jacob Feldman, DecisionCAMP Chair (Slides Recording)
- Denis Gagne (Trisotech): How do you FEEL about low code? (Slides Recording)
- Greger Ottosson (IBM): Transparent Machine Learning for Data-driven Decision Management (Slides Recording)
- Alan Fish (FICO): Decision Models for Ethical Decision Making (Slides Recording)
- KEYNOTE: Paul Vincent (Industry Analyst, Gartner): The Evolution of Decisioning in IT, and What Happens Next (Slides Recording)
- Bob Moore (JETset Business Consulting Ltd): Managing the Business Data in Business Decisions (Slides Recording)
- Seth Meldon (Progress): Framework Agnostic, Rule-Driven Dynamic Forms (Slides Recording)
- Silvie Spreeuwenberg (Simacan): Always look at the bright side (Slides Recording)
- Chris Berg (inrule): Good Beginnings – Sustainable Automation (Slides Recording)
- Rob Parker (Angle Finance): Engineering Decisions for Decision Engineering (Slides Recording)
- Jan Vanthienen (KU Leuven): You built a decision model! Now what? (Slides Recording)
- Jacob Feldman (OpenRules): Decision Modeling: Good, Bad, Ugly (Slides Recording)
- Ron Ross (Business Rule Solutions): Decision Tables: Could They Communicate Better? (Slides Recording)
- KEYNOTE: Christian De Sainte Marie (Program Manager at IBM): Neuro-Symbolic AI and Decision Rules (Recording)
- Bruce Silver (Bruce Silver Associates): The Killer App for DMN (Slides Recording)
- Denzil Wasson and Larry Goldberg (Sapiens Decision): Knowledge Driven Decision Management (Slides Recording)
- Daniel Schmitz-Hübsch (Materna Information & Communications SE): Introduction of DMN in projects: A marathon and not a sprint (Slides Recording)
- Guilhem Molines (IBM) and Don Scott (Marine AI): Marine Autonomy and Decisions (Slides Recording)
- Wai Wong, Joost Vennekens, Walter Schaeken, Lorenz Demey (KU Leuven): Deontic DMN: Representing deontic statement in DMN with bitstring semantics (Slides Recording)
- Greger Ottosson (IBM): Explaining Decisions made with both Business Rules and Machine Learning (Slides Recording)
- KEYNOTE: Ian Horrocks (University of Oxford): Knowledge Graphs: Theory, Applications and Challenges (Recording)
- Davide Sottara, Jane Shellum, Adam Bartscher and Deborah Sita (Mayo Clinic): Data Enriched Knowledge at your fingertips: delivering contextualized Clinical Decision Support (Slides Recording)
- Denis Gagne (Trisotech), Tom DeBevoise (Advanced Component Research): Intelligent Assistance for Knowledge Workers (Slides Recording)
- Carole-Ann Berlioz (Sparkling Logic): A practical guide to a decision migration project (Slides Recording)
- Russell McDonell (C-Cost System Pty Ltd): Is Python a good fit for DMN? An exploration of a Python implementation (Slides Recording)
Interactive Discussion “Ask an Expert”: Sep 27, 16:00 EST (moderated by James Taylor, Decision Management Solutions)
- Carlos Serrano-Morales, Sparkling Logic
- Jan Purchase, Lux Magi
- Guilhem Moliness, IBM
- Bob Moore, JETset Business Consulting
- Alan Fish, FICO
- Gary Hallmark, Oracle
Presentation Abstracts and Author Biographies
How do you FEEL about low code? by Denis Gagne (Trisotech)
Digital transformation and the accelerated transition to remote work are contributing to a perfect technological storm. This perfect storm is indifferently hitting every industry around us. A particularly challenging vector of this technological storm is the ever-growing need for business automation to achieve digital transformation conflated with an ever-growing shortage of technology professionals and software developers. With most companies turning to technology to transform how they engage with customers; software developers are in high demand and short supply. It is clear that we cannot rely on this small number of specialised workers – software developers – to carry out the massive undertaking of digital transformation in organizations.
One way to weather the storm is to empower non-developers in organizations to automate business logic. Business knowledge workers within organizations have a clear understanding of the logic of the business. They have the best understanding of business workflows and decisions required to deliver and exceed the new and expected digital customer experience. They excel at business decision thinking. Then why not enable these business knowledge workers to become not only the business logic architects, but also the actual construction workforce of your digital transformation?
In this session we discuss the emergence of the low-code paradigm as a required enabler to the timely achievement of the desired digital transformation. We compare the notions of no-code, low-code and pro-code and discuss how business knowledge workers can learn to think more like software developers by adopting a Decision Thinking mindset. Using the Decision Model and Notation (DMN) as the cornerstone of decision thinking and the Friendly Enough Expression Language (FEEL) as the low-code language of choice, we show how business knowledge workers can take business automation to production faster, gain simple and efficient ways of making enhancements, and maintain the deployed automated business logic. FEEL is simple enough for business knowledge workers yet expressive and powerful enough for professional developers. In short, FEEL offers the perfect scaffolding for the automation of business logic. With FEEL as a low code language, business knowledge workers can truly become the artisans of the digital transformation. Keywords: Low-Code, Decision Thinking, Friendly Enough Expression Language (FEEL), Decision Model and Notation (DMN)
For over a decade Denis Gagné has been a driving force in the majority of international BPM standards in use today. He is a member of the Workflow Management Coalition (WfMC) Steering Committee, chair of the Business Process Simulation Working Group (BPSWG), and co-Editor of the XPDL 2.2 process definition standard. For the Object Management group (OMG), Denis is the Chair of the BPMN Interchange Working Group (BPMN MIWG), and a member of the Business Process Model and Notation (BPMN), Case Management Model and Notation (CMMN) team and Decision Management (DMN) team.
Transparent Machine Learning for Data-driven Decision Management by Greg Ottosson (IBM)
As businesses strive to become more “data-driven” and make more precise predictions with Machine Learning, one emerging issue is transparency. For critical decisions, both internal business guidelines and external regulation might require our ML-based predictions to be as transparent as the existing business rules they complement. While black-box ML models can be leveraged – perhaps with post-hoc explainability added – in many cases a more suitable approach might be to train a fully transparent ML model.
Rule Sets and Score Cards have been hand-crafted and used in Decision Management for decades but deriving them automatically from data creates new opportunities and some new challenges. How should these transparent ML models be included in decision models? How should they be expressed, managed, and governed? How should they be invoked and explained in the context of additional policy rules?
In this presentation we discuss the characteristics and use cases of transparent ML models, and how they can be included inside decision models, while enhancing both accuracy and governance of decisions.
Keywords: Machine Learning, Rule Learning, Decision Modeling, Predictive Analytics, Transparency
Decision Models for Ethical Decision Making by Dr. Alan Fish (FICO)
Around the world, consumers are becoming increasingly concerned about the ethicality of automated decisions. This presentation will suggest ways in which decision models can be used to ensure that decisions are transparent, explainable and human-centered. Keywords: DMN, Decision Model, Decision, Ethics, Explanation.
Dr. Alan Fish is an authority in Decision Modelling and Decision Management, especially in the support and/or automation of organisational decision-making. With over 30 years experience in this field, he has been responsible for many projects at the forefront of current technology. He invented the “Decision Requirements Diagram” (DRD) which exposes the structure of a domain of decision-making, and developed Decision Requirements Analysis (DRA): a methodology for building and using such decision models. He is the author of “Knowledge Automation: How To Implement Decision Management in Business Processes” (Wiley), and a co-author of the OMG specification Decision Model and Notation (DMN).
Making decisions involves three distinct bits of knowledge: the case data which identifies the specific situation we are making a decision on; the logic behind how we make decisions; and the internal ‘business data’ items which set the thresholds and constraints the logic applies to the case data. Such business data appears in the cells of our decision tables and in expressions in our rules and functional logic.
Often the number of business data items involved may be small, only a few dozen if that. But in some cases, there may be hundreds, thousands, even millions of such items. Some of this data may also be only partially under business control, being provided by third party regulatory or standards bodies. === Even with only a few data items to manage, problems can arise. How do we efficiently represent how the data drives the decisions? Will our decision systems give timely results if we potentially have scan thousands or even millions of data items? How is maintenance performed as the business data changes over time?
There is no one size fits all solution to these issues, but there are many strategies one can employ to incorporate data into decisions. The presentation will explore a small subset of these mainly using techniques explored in some of the Challenges set by the Decision Management Community in recent years, but ones which have also been exploited in real world applications, in finance, insurance and manufacturing among other industry sectors. Keywords: Business Data, Decision Management, Implementation Strategies
Dr. Bob Moore started work on automated decision systems in 1989 and spent the following three decades or so building rule-based systems for a range of clients in finance, insurance, manufacturing and government, using several different technologies. He is largely retired now but has recently got himself involved in educating the next generation of IT professionals as an Associate Lecturer in the School of Computing and Communications at the Open University.
Framework Agnostic, Rule-Driven Dynamic Forms by Seth Meldon (Progress)
Seth Meldon is the senior solution engineer supporting the Corticon business rules engine in North America at Progress, supporting new and existing users throughout all phases of designing rule-intensive applications. Melding this professional experience with his areas of personal research interest, he has designed and written about solutions leveraging the FHIR standard to improve patient access to their medical data, implementing decision logic for incentive-based community adoption of renewable energy sources, and running decision services as cloud-based functions and in mobile applications. His independent, serial investigative articles and longform environmental history report, The Watershed, are accessible from sethmeldon.com
Always look at the bright side by Silvie Spreeuwenberg (Simacan)
As an advocate of business rules engine technology and entrepreneur in the software industry, with a background in Artificial Intelligence, I ended up recently as CTO of a tech scale-up. With 7 year old high performing, high code software the stack of backlog features is getting higher every day and
the pullboard size seems to become smaller every iteration. We follow the Agile Methodology based on SCRUM and some aspects of SAFE but agility is far away. Sounds familiar?
Introducing a business rule engine to externalize the management of business logic and decrease the demand for coding, is a major step in such environment. A smaller step, such as a low code platform for green field features, is just as challenging. I share with you the obstacles I encountered and we brainstorm solutions.
Keywords: Business Rules, Low Code, No Code, Agile software development, Case Study, Lessons Learned, Operational business decision services
Good Beginnings – Sustainable Automation by Chris Berg (inrule)
Somewhere in the life of an automation project a transition takes place from surviving to thriving. The first go-live will feel much different after two hundred promotions to production. Now the organization must live with the solution. Living with automation requires the same rigor and careful execution as the initial phases of the project. What’s in the backlog after go-live? We find a list of important deferred features, bugs, and technical debt. Many tasks remain manual and while the finish line approached, the team expended little toward automating the lifecycle or tracking data for business impact. Long-term success requires a broad view of the investment across teams and activities making sure the effort doesn’t drift or get mired in manual activities. Most importantly the team must finish the work required to confirm the business case.
Keywords: team building, core values, design thinking, strategy, practice
Chris Berg Director, Product Strategy and Design Chris has been leading people, products, design and technology for over 15 years in the enterprise software space. He has led significant design explorations (BPM, PaaS, API Management and DevOps) while at IBM and covers more than 10 years of product leadership in the Decision Management space. In much of this time, he focused on transforming business behavior and empowering business users. In his role at InRule Technology, Chris is responsible for product strategy, design and alignment with the market.
Engineering Decisions for Decision Engineering by Rob Parker (AngleFinance)
In this presentation we discuss our experiences whilst implementing a credit policy from text based policy rules to a Decision Model Notation (DMN) implementation. In particular, we focus on the engineering decisions or implementation design decisions necessary to implement a transparent, maintainable decision automation solution. Engineering design choices include; the granularity of decisions; when to choose between FEEL expressions versus decision tables; when to use compound business objects (json structures) versus when to use primitive types as inputs to decisions; how much FEEL is too much FEEL. In addition we discuss design choices such as the impact the user experience may have on the structure and sequencing of decisions and vice versa. Keywords: DMN, FEEL, Decision Automation, Credit Policy, Engineering
You built a decision model! Now what? by Jan Vanthienen (KU Leuven)
DMN offers a business-friendly representation of business decision knowledge and current tools use this executable specification to make a decision based on this knowledge. DMN even allows checking the model for consistency and completeness. However, there is more than straightforward processing of input cases.
DMN also allows a number of other functionalities that go beyond simple (or complex) decisioning:
– flexible reasoning, explanation, answering decision questions like How, Why, What-If, How-To (e.g. in the form of a chatbot), as presented at last year’s DecisionCamp
– decision analysis: what are the most important criteria?, what is the average cost of a decision?, what are the most frequent cases? Who is doing what?
– fairness: does the knowledge correspond to what one would expect in terms of changes in information item values. E.g. does a higher income always lead to a higher eligibility?
– compliance: is the decision knowledge compliant with existing rules and regulations?
– monitoring: how many cases actually obtained a specific decision outcome?
– evaluation of the decision policy based on historical data
– simulation and prediction: What would be the total outcome of a policy change?
This presentation provides an overview of opportunities for decision modeling and DMN by looking into some common business challenges.
Keywords: Decision Model and Notation, DMN, decision modeling, decision analysis
Prof. Jan Vanthienen received his PhD degree in Applied Economics from KU Leuven, Belgium. He is a full professor of Information Systems at the Department of Decision Sciences and Information Management, KU Leuven and (co-)authored more than 200 full papers in international journals and conference proceedings. His research interests include information and knowledge management, business rules, decisions and processes, and business analysis and analytics. He received an IBM Faculty Award on smart decisions, and the Belgian Francqui Chair at FUNDP. Currently he is department chair at the Department of Decision Sciences and Information Management of KU Leuven.
Decision Modeling: Good, Bad, Ugly by Dr. Jacob Feldman (OpenRules)
In this presentation I will share different implementations of the decision model for the recent DM Community Challenge “Medical Claim Processing“. I will show how a desire to use already known decision modeling constructs could lead to bad (and even ugly) decision model implementations. We will discuss certain pitfalls, and I will use this example to provide practical advices for building intuitive and manageable decision models for complex business problems.
Keywords: Business Decision Models, Decision Microservice, Invocation Context, Authorized access, Security, CI/CD
Dr. Jacob Feldman is the CTO of OpenRules, Inc., a US corporation that created and maintains the highly popular Business Rules and Decision Management System commonly known as “OpenRules”. He has extensive experience in development of decision-making engines using business rules, optimization, and machine learning technologies for real-world mission-critical applications. Jacob is the DecisionCAMP’s Chair, the manager of DMCommunity.org, and an active contributor to BR&DM forums. He is also the Specification Lead for the optimization standard JSR-331. Dr. Feldman is an author of two books “DMN in Action with OpenRules“ and “Goal-Oriented Approach to Decision Modeling“. He has 5 patents and many publications in the decision intelligence domain.
Decision Tables: Could They Communicate Better? by Ronald G. Ross (Business Rule Solutions)
Are decision tables computational devices or communication devices? Yes (both!). What do communities of people need for effective communication? Meaning. Sense-making is a key element of all communication. Decision tables (and decision rules) are no exception. So, the vocabulary used in decision tables should be defined and structured.
And why do many decision tables become so complicated? Perhaps because exceptions and scope are not being handled as well as they could be.
There’s a whole world beyond decision tables. Decision tables should make a better effort to communicate with it. This presentation includes real-world examples of good and not-so-good communication to stimulate discussion.
• Understanding the Elements of Vocabulary Structure
• Specifying Applicability Explicitly
• Behavioral Rules as Guardrails
Keywords: Decision Tables, Rules, Vocabulary, Exceptions, Scope, SBVR
Ronald G. Ross is Co-Founder and Principal of Business Rule Solutions, LLC (www.BRSolutions.com). BRS provides consulting, training and mentoring in support of policy analysis, business rules, concept modeling, decision analysis, and business knowledge engineering. BRS clients have included many 100s of top businesses and government bodies world-wide. Ron is the author of the 2020 groundbreaking book “Business Knowledge Blueprints: Enabling Your Data to Speak the Language of the Business”, featuring concept models, business vocabularies and disambiguation. It is his 9th professional book.
Ron is Chair of the annual Building Business Capability (BBC), official conference of the IIBA®. He is also Executive Editor of BRCommunity.com and its flagship on-line publication, Business Rules Journal. Ron has keynoted dozens of conferences and given seminars to many thousands of people worldwide. Ron co-develops the landmark BRS methodology featuring numerous innovative techniques including the popular RuleSpeak® (free on RuleSpeak.com). These are the latest offerings in a 45-year career that has consistently featured creative, business-driven solutions. Ron is recognized internationally as the ‘father of business rules.’ In 2017 he was co-author with John Zachman and Roger Burlton of the Business Agility Manifesto (www.busagilitymanifesto.org). Mr. Ross holds an M.S. in information science from the Illinois Institute of Technology and a B.A. from Rice University. .
The Killer App for DMN by Bruce Silver (Bruce Silver Associates)
When I introduced DMN to DecisionCamp in 2016, the standard was brand new. It promised something unusual for a standard: Beyond a way to create decision requirements, DMN gave business users the ability to create executable decision logic themselves, without programming, using tabular formats called boxed expressions and a friendly expression language FEEL for the table cells. By 2017, Red Hat had provided an open source runtime for it, but most vendors claiming DMN support still neglected FEEL and boxed expressions.
Fast forward to today: Although the posture of Decision Management vendors is largely unchanged, the world of software has changed. The cloud has transformed it completely. REST APIs are everywhere. And you can’t hire skilled programmers for love or money. As a result, the demand for Low-Code Business Automation is exploding, and incorporating boxed expressions and FEEL into BPMN offers a standards-based platform for that.In this presentation, we’ll see how that works on the Trisotech platform, where DMN is used for all modeler-defined business logic – not just decisions – and for mapping between process variables and the REST APIs of essentially all process tasks. Now subject matter experts can create complex Business Automation solutions themselves, without programming, and deploy them in one click as cloud REST services.
There are many Low-Code platforms today based on proprietary script. FEEL and boxed expressions are not only non-proprietary but more powerful and more business-friendly. Given the stubborn resistance of DM vendors, the real opportunity for DMN is not Decision Management but Low-Code Business Automation. Keywords: DMN, Low-Code, Business Automation
Bruce Silver is a well-known consultant, industry analyst, and educator specializing in BPM. He is founder and Principal of BPMessentials, the world’s leading provider of BPMN training and certification, and methodandstyle.com. Author of BPMN Method and Style and DMN Method and Style, Dr Silver served on the OMG technical committees that developed the BPMN 2.0 and DMN 1.1 standards. Previously, he served on the board of directors of Captiva Software until its acquisition by EMC in 2005, and was Vice President and head of workflow and document management research at the industry analyst firm BIS Strategic Decisions (which became Giga Information Group, now Forrester Research). In the 1980s, he was engineering manager at Wang Labs in charge of one of the very first commercial document imaging and workflow systems. He holds Physics degrees from Princeton and MIT, and four patents in imaging and workflow.
Knowledge Driven Decision Management by Denzil Wasson and Larry Goldberg (Sapiens Decision)
Case study of the use of knowledge and decision management allowing business analysts to provision business solutions. We introduce the concept of a knowledge/decision driven method to model and produce event-based architecture solutions. This approach effectively externalizes business logic and the surrounding use cases to create situationally aware solutions that respond to change in the hands of the business.
Sectors: Banking/Mortgage (Demo use case) – approach and technologies are sector neutral.
Technologies: Sapiens Knowledge, Sapiens SAFr, Sapiens Decision. Keywords: Knowledge, Decision management, Event-based architecture, Situationally aware solutions
Introduction of DMN in projects: A marathon and not a sprint by Daniel Schmitz-Hübsch (Materna)
In the past years the DMN standard has been further developed. Among other things, attention has been paid to improving the user-friendliness of decision modeling. At the same time, the DMN standard was increasingly introduced in customer projects. Especially the balancing between the acceptance of the business side and the technical implementation has been a challenge to break resistance.
The presentation will recapitulate how the acceptance of DMN has developed in various real-world projects over the last years and how the adaptations and improvements of the DMN standard have been received. Furthermore, the tools that have proven useful in the requirements engineering will be presented. However, the development of the DMN standard and the tools is not yet complete. At the end of the presentation, concrete suggestions for improvement of the DMN standard as well as the tools based on the project experiences will be discussed to further increase the use and acceptance of DMN. Keywords: DMN, FEEL, Rule Engines
Daniel holds a Master degree in Business Informatics with focus on mobile oriented analysis of business processes. For eight years, he has been involved in the modelling and technical implementation of business process- and decision management systems. As a software developer for an independent IT company, he is responsible for the development of high-availability decision applications using rule engines like IBM Operational Decision Management and RedHat Drools.
Marine Autonomy and Decisions by Guilhem Molines and Don Scott (IBM)
Mayflower Autonomous Ship is a custom-built vessel which has recently crossed the Atlantic ocean with only software and no human at the helm. Designed by UK-based Marine AI, with IBM software, the AI Captain in charge of the navigation combines various AI components, including Machine Learning, Decision and Optimization Models.
This session presents the vision for the project, and will focus on the technology in use to take decisions in this hostile environment with no human intervention. The talk will also give an update on the whereabouts of the ship. Keywords: AI, Decision, Autonomy, Marine
Deontic DMN: Representing deontic statement in DMN with bitstring semantics by Wai Wong, Joost Vennekens, Walter Schaeken, Lorenz Demey (KU Leuven)
Decision Model and Notation (DMN) is a standard for modelling operational decisions. It is intended to be used in an industrial setting by domain experts, who may not have a background in logic. In many settings, the decision processes that need to be modelled refer to deontic concepts, such as obligation, prohibition and permission. However, since DMN does not contain such concepts, domain experts must model them implicitly. This may lead to problems, for example, because of an inability to distinguish deontic inconsistency from definitional inconsistency. To solve this issue, we propose to extend DMN with deontic concepts. We use bitstring semantics for this, a recent logical framework developed in logical geometry. In this way, we can nicely incorporate the deontic hexagon, a logical diagram that captures the relations among different deontic concepts, into DMN.
In this talk, we will present this deontic extension of DMN and its advantages and challenges. For example, one advantage is the possibility to combine multiple deontic decisions rendered by different decision tables into a single output column according to deontic logic, as opposed to multiple output columns in traditional DMN. Meanwhile, one example challenge would be that humans interpret deontic terms differently than how it is defined in logic (e.g., a permission is often interpreted as “permitted but not obligated” instead of the logical “permitted and possibly obligated”), a phenomenon known as scalar inference. We will discuss our approaches in solving those challenges and their potential implementations with our reasoning engine e.g. IDP-Z3. Keywords: DMN, Bitstring semantics, Deontic logic, Scalar inference
Joost Vennekens is an associate professor at KU Leuven Campus De Nayer in Sint-Katelijne-Waver, Belgium. His research is concerned with AI technology (both Knowledge Representation and Machine Learning) and its industrial applications. He belongs to the research group EAVISE, which focuses on AI, computer vision and embedded systems, and to the research group DTAI, which studies declarative languages and AI. He is a member of the board of the Benelux Association for Artificial Intelligence and of the board of the Leuven.AI institute.
Explaining Decisions made with both Business Rules and Machine Learning by Greger Ottosson (IBM)
Historically, we’ve considered decision management as intrinsically explainable. Through a “rule trace” we can introspect and understand why a decision model produces a specific outcome. However, while a trace might be sufficient to allow the rule author to debug a model, this procedural understanding doesn’t lend itself well to explain the decision to the end user (such as the recipient of a mortgage approval, product recommendation or fraud alert). It is therefore not uncommon to construct a tailored explanation as a side-effect of the actual decision, such as the “reason codes” common with rules and score cards. While reason codes are helpful, constructing and aggregating them requires work and they don’t directly provide information about what would need to change in order to arrive at a different decision – a so-called “counterfactual” explanation.
While explaining decisions made with rules are hard enough, the inclusion of Machine Learning models adds additional complexity. Unless these ML models are constrained to transparent types – such as smaller rule sets, decision trees or score cards – we’re are now faced with explaining (perhaps multiple) black-box ML models within the context of business rules. While it’s possible to treat the entire decision as a black-box model, ignoring the declarative information present in the business rules can lead to confusing explanations. In this presentation, we will present new research with the long-term goal of providing a unified approach to explaining decisions based on both ML-based predictions and business rules – providing both attributions to input variables and counterfactual explanations.
Keywords: Machine Learning, Explainability, Explainable Decisions
Data Enriched Knowledge at your fingertips: delivering contextualized Clinical Decision Support by Davide Sottara, Jane Shellum, Adam Bartscher and Deborah Sita (Mayo Clinic)
Most Clinical Decision Support (CDS) Systems used for actual patient care are based on Event-Condition-Action (ECA) Rules. Standards such as SMART on FHIR and CDS Hooks are the de facto way to access the necessary patient data. Various forms of AI, including decisions, rules, and predictive models, formalize when, whether and how to process the data to derive insights and recommendations. Comparatively, there are fewer standard methodologies to deliver the actual inferences back to patients and/or care providers, especially when explanations, background knowledge and contextualization are necessary to build trust and illustrate the proposed course of action.
To this end, we propose an approach based on the contextualization of intelligent content. Editors with clinical expertise curate the background medical knowledge and expert insights as semi-structured, modular content using the Darwin Information Typing Architecture (DITA) standard. Knowledge Engineers augment the content with semantic annotations, to formalize aboutness, applicability and contextualizable parameters. DITA’s built-in pattern matching mechanism allows to assemble, filter, (de)emphasize and/or specialize the content, using features of the patient case, derived using a combination of inference techniques.
We have further established a predicate logic interpretation of the contextualization rules, which supports the implementation of a real-time templating engine, but also enables the use of techniques such as consistency checking and satisfiability to verify the correctness and completeness of the content with respect to the intended scenarios.
We have implemented the framework as part of a cognitive support application, which delivers care guidance for Mayo Clinic patients with cardiac conditions.
Keywords: Clinical Decision Support, DITA, Intelligent Content, Decision Models
Intelligent Assistance for Knowledge Workers by Denis Gagné (Trisotech) and Tom DeBevoise (Advanced Component Research)
Knowledge Workers, a term coined by Peter Drucker, are workers whose job is to think for a living. Knowledge work can be differentiated from other forms of work by emphasizing continuously evolving non-routine problem-solving based on information. As businesses increase their dependence on information technology via digital transformation, the number of fields in which knowledge workers must operate has expanded dramatically.
Today, much of the knowledge work accomplished involves informal collaborations via emails supported by attached documents (PDFs and others). Fundamentally, knowledge workers spend much of their time acting as human integrators of unstructured information exchanged via unstructured communications and collaborations.
In support of these efforts, Intelligent Document Process (IDP) technologies were introduced by various vendors to transform unstructured and semi-structured information into usable data. The ultimate objective of most IDP capabilities is to integrate with downstream systems such as ERP. They tend to be based on pattern matching supported by Machine Learning (ML) technologies. To become effective, these approaches require varying quantities of representative information being available or supervised learning and labeling techniques that is yet another form of knowledge work. But what if an adequate sample of examples or information are not available for a particular type of knowledge work? And how do we support knowledge workers and their actual flow of work?
In this session, we will present a combination of symbolic and non-symbolic reasoning techniques to ease the burden on knowledge workers by offering intelligent just-in-time assistance. This approach is based on open international workflow and decision standards and anchored on the low-code Friendly Enough Expression Language (FEEL) from the Decision Model and Notation (DMN). We use Natural Language Processing (NLP) to enable knowledge-based workflows with channels of intelligent email messages. NLP detection, mediated by decision models of email-created events triggers the flow of knowledge work, detects intermediate business events, route attachments and results for approval or exceptions, and provides useful information to knowledge workers, including calendar events, contacts, and various reports. A Real Estate Closing Process will be used as an example. Keywords: Knowledge Work, Intelligent Document Process (IDP), Natural Language Processing (NLP), Decision Model and Notation (DMN), Friendly Enough Expression Language (FEEL)
A practical guide to a decision migration project by Carole-Ann Berlioz (Sparkling Logic)
Abstract..Keywords: Credit Origination, Financial Services, Business Rules and Decision Modeling, Case Studies / Lessons Learned, Best Practices, Design Patterns, Tips & Tricks
Carole-Ann is co-founder and CPO of Sparkling Logic, a leader in Decision Management technology. Over her 15+ years in Decision Management, she has consistently brought innovation in the business rules and decision management space, which has been recognized by leading industry analysts Gartner and Forrester. She holds several patents in decision management and adaptive modeling. Carole-Ann started her career building Expert Systems and Business Intelligence dashboards, then specialized in Business Rules / Optimization at ILOG and finally led the vision and direction for Blaze Advisor & Decision Management tools at FICO. She is a passionate & renowned blogger and speaker.
Is Python a good fit for DMN? An exploration of a Python implementation by Russell Mc Donell (C-Cost System Pty Ltd)
You could also ask “Is JSON a good fit for DMN?” and the answer, from any API developer, would be “It had better be or DMN is dead in the water!”. Turns out DMN and JSON play well together. And we know that Python has good JSON support, so Python/DMN/JSON should be a good combination.
Why? Well, DMN is based upon FEEL – a loosely typed expression language. JSON is a loosely typed data structure. And Python is a loosely typed programming language. easier. Python supports all the FEEL data types, except ‘year and month duration’. So, for the most part, implantation should be easy. All we need is a FEEL parser. Python has the ‘SLY’ module – a lex and yacc parser. With SLY it is possible to implement a FEEL parser (pySFeel) in under 4,000 lines of code.
How? An implementation of DMN has to read in a DMN specification. Here there are lots of options. The obvious staring point was the pretty pictures that people can see in the DMN specification, those decision tables. And the easiest and cheapest way for users to create those pretty pictures is Excel. And Python has ‘openpyxl’; an excellent module for reading in, and parsing, Excel workbooks. It’s possible to search an entire Excel workbook and find all the DMN compliant decision tables. Which makes it possible to implement a DMN rules engine (pyDMNrules) in under 4,500 lines of code.
Keywords: Python, Excel, pyDMNrules, pySFEEL