Decision Making in AI

Good thoughts here ... its not all about decisions, but decisions in real contexts. 

The Future of Decision-Making in AI

How not to end up in a Skynet, The Architect, Ultron or HAL9000 situation

By Alex Elkjær Vasegaard

 Generally, I am not a fan of the terminology Artificial intelligence (AI). It is too broad, and non-technical minded people imagine that AI is a singular entity that makes decisions independently. Additionally, because AI is a popular term, I have seen examples where companies advertise themselves using AI when they are actually “just” using linear regression. Throughout the last 80 years, the term has gotten a bad rap in pop culture because of all the doomsday science-fiction stories and movies. Countless times we have seen science-fiction turning into science-faction, and with the advent of the text generator GPT-3 by OpenAI, it sure looks like we are on track. So, will this also happen here?

Nope.  .... ' 

Responsible and Explainable AI

I would call this more broadly, 'responsible decision making' to embrace what is called 'AI' and analytical decision making.  Useful and considered look at what this means. 

Responsible AI: Bridging From Ethics to Practice   By Ben Shneiderman  via CACM

Communications of the ACM, August 2021, Vol. 64 No. 8, Pages 32-35  10.1145/3445973

The high expectations of AI have triggered worldwide interest and concern, generating 400+ policy documents on responsible AI. Intense discussions over the ethical issues lay a helpful foundation, preparing researchers, managers, policy makers, and educators for constructive discussions that will lead to clear recommendations for building the reliable, safe, and trustworthy systems6 that will be commercial success. This Viewpoint focuses on four themes that lead to 15 recommendations for moving forward. The four themes combine AI thinking with human-centered User Experience Design (UXD).

Ethics and Design. Ethical discussions are a vital foundation, but raising the edifice of responsible AI requires design decisions to guide software engineering teams, business managers, industry leaders, and government policymakers. Ethical concerns are catalogued in the Berkman Klein Center report3 that offers ethical principles in eight categories: privacy, accountability, safety and security, transparency and explainability, fairness and non-discrimination, human control of technology, professional responsibility, and promotion of human values. These important ethical foundations can be strengthened with actionable design guidelines.

Autonomous Algorithms and Human Control. The recent CRA report2 on "Assured Autonomy" and the IEEE's influential report4 on "Ethically Aligned Design" are strongly devoted to "Autonomous and Intelligent Systems." The reports emphasize machine autonomy, which becomes safer when human control can be exercised to prevent damage. I share the desire for autonomy by way of elegant and efficient algorithms, while adding well-designed control panels for users and supervisors to ensure safer outcomes. Autonomous aerial drones become more effective as remotely piloted aircraft and NASA's Mars Rovers can make autonomous movements, but there is a whole control room of operators managing the larger picture of what is happening.

Humans in the Group; Computers in the Loop. While people are instinctively social, they benefit from well-designed computers. Some designers favor developing computers as collaborators, teammates, and partners, when adding control panels and status displays would make them comprehensible appliances. Machine and deep learning strategies will be more widely used if they are integrated in visual user interfaces, as they are in counterterrorism centers, financial trading rooms, and transportation or utility control centers.

Explainable AI (XAI) and Comprehensible AI (CAI). Many researchers from AI and HCI have turned to the problem of providing explanations of AI decisions, as required by the European General Data Protection Regulation (GDPR) stipulating a "right to explanation."13 Explanations of why mortgage applications or parole requests are rejected can include local or global descriptions, but a useful complementary approach is to prevent confusion and surprise by making comprehensible user interfaces that enable rapid interactive exploration of decision spaces.   ... '

Note in particular a look at 'trustworthy certification' proposed outlines for a number of large industries, to show how this might be applied by the nature of their operation ... '

Bayer's use of Digital Twins, Decision Science

Another example of AI and Digital Twins being used to construct virtual models at major enterprises.

Bayer uses digital twins to reshape business strategy
Bayer Crop Science has created 'virtual factories' to provide dynamic digital representations of the equipment and processes for each of its nine North American corn seed manufacturing sites.
     By  Thor Olavsrud in CIO

For the past several years, Bayer Crop Science has been working to embed decision science into every facet of its business, from logistics to genetic sequencing. The strategy, driven by machine learning and artificial intelligence, now includes digital twins, or "virtual factories," that model each of the nine corn seed manufacturing sites that Bayer operates in North America.   ... " 

Mixing Machine Learning and Human Judgement

Ultimately this will be a very important problem.  We worked on many examples where we could create very complex analytics or AI style solutions, but ultimately fit their solutions into real world decisions.    What is the form of the mixing?  Assistant, Simulation, Testing, Enhancing, Collaboration?   The form of mixing will be key.

The Effects of Mixing Machine Learning and Human Judgment

Collaboration between humans and machines does not necessarily lead to better outcomes.
Michelle Vaccaro and Jim Waldo  in Queue

In 1997 IBM's Deep Blue software beat the World Chess Champion Garry Kasparov in a series of six matches. Since then, other programs have beaten human players in games ranging from Jeopardy to Go. Inspired by his loss, Kasparov decided in 2005 to test the success of Human+AI pairs in an online chess tournament.2 He found that the Human+AI team bested the solo human. More surprisingly, he also found that the Human+AI team bested the solo computer, even though the machine outperformed humans.

Researchers explain this phenomenon by emphasizing that humans and machines excel in different dimensions of intelligence.9 Human chess players do well with long-term chess strategies, but they perform poorly at assessing the millions of possible configurations of pieces. The opposite holds for machines. Because of these differences, combining human and machine intelligence produces better outcomes than when each works separately. People also view this form of collaboration between humans and machines as a possible way to mitigate the problems of bias in machine learning, a problem that has taken center stage in recent months.12

We decided to investigate this type of collaboration between humans and machines using risk-assessment algorithms as a case study. In particular, we looked at the COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) algorithm, a well-known (perhaps infamous) risk-prediction system, and its effect on human decisions about risk. Many state courts use algorithms such as COMPAS to predict defendants' risk of recidivism, and these results inform bail, sentencing, and parole decisions.

Prior work on risk-assessment algorithms has focused on their accuracy and fairness, but it has not addressed their interactions with human decision makers who serve as the final arbitrators. In one study from 2018, Julia Dressel and Hany Farid compared risk assessments from the COMPAS software and Amazon Mechanical Turk workers, and found that the algorithm and the humans achieved similar levels of accuracy and fairness.6 This study signals an important shift in the literature on risk-assessment instruments by incorporating human subjects to contextualize the accuracy and fairness of the algorithms. Dressel and Farid's study, however, divorces the human decision makers and the algorithm when, in fact, the current model indicates that humans and algorithms would work in tandem. .... " 

O'Reilly Data Show Podcast: Linking Decisions and Data Science

Liked this very much,  in particular the statement of the topic ... that its all about how we link data science to real life decisions.   The need for  'Decision Science', and social and behavioral understanding in context.   Yes, please.  Following this podcast now, available on most podcast services, I had it up in seconds on Alexa via Tunein.

Make data science more useful
The O’Reilly Data Show Podcast: Cassie Kozyrkov on connecting data and AI to business.

By Ben Lorica August 1, 2019

Cassie Kozyrkov will deliver a keynote on safe, reliable, and responsible machine learning at the Strata Data Conference in New York City, September 23-26, 2019. Early price ends August 9.

Subscribe to the O'Reilly Data Show Podcast to explore the opportunities and techniques driving big data, data science, and AI. Find us on Stitcher, TuneIn, iTunes, SoundCloud, RSS.

In this episode of the Data Show, I speak with Cassie Kozyrkov, technical director and chief decision scientist at Google Cloud. She describes "decision intelligence" as an interdisciplinary field concerned with all aspects of decision-making, and which combines data science with the behavioral sciences. Most recently she has been focused on developing best practices that can help practitioners make safe, effective use of AI and data. Kozyrkov uses her platform to help data scientists develop skills that will enable them to connect data and AI with their organizations' core businesses. ... "

See also:   Cassie Kozyrkov blog   https://medium.com/@kozyrkov 

Great Decision Making

Good piece which get to the point of decision and process.  And also ultimately in AI.  The solution of commuting to a default decision up front is interesting, never seen it implemented.   And would seem to depend on the risk of alternate decisions,  and understanding them in alternate contexts.  Seeking data in alternate contexts is commendable, but often hard.

The First Thing Great Decision Makers Do
By Cassie Kozyrkov in HBR

As a statistician, I appreciate the quote by applied statistics pioneer W. Edwards Deming, “In God we trust. All others bring data.” But as a social scientist, I’m compelled to warn you that many decision-makers chase data with too much zeal, running from ignorance but never improving their decisions. Is there a way to land in the sweet spot? There is, and it starts with one simple decision-making habit: Commit to your default decision up front.

The key to decision-making is framing the decision context before you seek data — a skill that unfortunately is not usually covered in data science courses. To learn it, you’ll need to look to the social and managerial sciences. It’s unfortunate that we don’t teach it enough where it is most needed: as a skill for leading and managing data science projects. Even in statistics, which is the discipline of making decisions under uncertainty, most of the exercises that students encounter already have the context pre-framed. Your professor usually creates the hypotheses for you and/or frames the question so there’s only one right answer. Wherever there’s a right answer, the decision-maker has already blazed that trail.

Many decision-makers think they’re being data-driven when they look at a number, form an opinion, and execute their decision. Unfortunately, such a decision will be “data-inspired” at best. Data-inspired decision-making is where we swim around in some numbers, eventually reach an emotional tipping point, and then decide. There were numbers near that decision somewhere, but those numbers didn’t drive the decision. The decision came from somewhere else entirely. It was there all along in the unconscious biases of the decision-maker.   .... " 

Seeking Great Data Analytics

Good piece, which even includes the rare 'decision' aspects included.    But implies the whole process is still very haphazard.  Suppose you wanted the automate the whole process, where would you start?

What Great Data Analysts Do — and Why Every Organization Needs Them
By Cassie Kozyrkov ... '

Better, Faster Decisions: Faster is a Context

Its all about decisions.   Not only our own but also those we make based on help from systems.   And, increasingly autonomous decisions.   And faster may not be better without understanding risk in various contexts.  And speed to applying an assisted decision is a context.

Three keys to faster, better decisions in McKinsey.   By Aaron De Smet, Gregor Jost, and Leigh Weiss

Decision makers fed up with slow or subpar results take heart. Three practices can help improve decision making and convince skeptical business leaders that there is life after death by committee.

Two years ago, we wrote about how it was simultaneously the best and worst of times for decision makers in senior management. Best because of more data, better analytics, and clearer understanding of how to mitigate the cognitive biases that often undermine corporate decision processes. Worst because organizational dynamics and digital decision-making dysfunctions were causing growing levels of frustration among senior leaders we knew.

Since then, we’ve conducted research to more clearly understand this balance, and the results have been disquieting. A survey we conducted recently with more than 1,200 managers across a range of global companies gave strong signs of growing levels of frustration with broken decision-making processes, with the slow pace of decision-making deliberations, and with the uneven quality of decision-making outcomes. Fewer than half of the survey respondents say that decisions are timely, and 61 percent say that at least half the time spent making them is ineffective. The opportunity costs of this are staggering: about 530,000 days of managers’ time potentially squandered each year for a typical Fortune 500 company, equivalent to some $250 million in wages annually. ... "

Intelligent Enterprise: Judgement, Reasoning and Decisions.

Stephen DeAngelis of Enterra Systems, who we have worked with writes on   The Rise of the Intelligent Enterprise  in LinkedIn: 

" ... We live in the Digital Age. The World Economic Forum has declared data is a resource as valuable as oil. We have watched the rise of digital enterprises (i.e., enterprises whose very existence was impossible until the Internet matured). Today most analysts agree organizations created in the Industrial Age need to undergo digital transformation and become digital enterprises. Some analysts even argue that going digital won’t be enough. To survive, they insist, an organization must become an intelligent enterprise. “In coming years,” explain Paul J.H. Schoemaker, Founder and Executive Chairman of Decision Strategies International, and Philip E. Tetlock (@PTetlock), the Annenberg University Professor at the University of Pennsylvania, “the most intelligent organizations will need to blend technology-enabled insights with a sophisticated understanding of human judgment, reasoning, and choice. Those that do this successfully will have an advantage over their rivals.”[1] .... " 

I add:

Good points about the understanding of human judgement, reasoning and choice.   Ultimately a result has to be inserted into decisions, usually a group of decisions, by a groups of people (or devices) over time, in varying contexts.   How do we continue to refine how decisions are made?

Analytics for Management

Good piece in the ACM, full text linked to below. And further it is not only about classical analytics but also about the emergence of cognitive aspects of AI in this space.    These approaches are more closely connected to the actual decisions that managers make. And how those decisions link together into a decision process.   We are not completely there yet, but approaching.  Management decisions are always a sequence of decisions, by multiple people,  in context.  That's not expressed enough in the below.

Analytics for Managerial Work  By Vijay Khatri, Binny M. Samuel 
Communications of the ACM, April 2019, Vol. 62 No. 4, Page 100
10.1145/3274277

A 2014 IDC report predicted that by 2020, the digital universe—the data we create and copy annually—will reach 44 zettabytes, or 44 trillion gigabytes.10 With the explosive growth in organizational data, there is increasing emphasis on analytics that can be used to uncover the "hidden potential" of data. A 2014 Society for Information Management survey found analytics/business intelligence to be #1 among the top 15 most significant IT investments in the prior five years.12 It is not surprising that business analytics is increasingly central to managerial decision making within business functions: finance, marketing, human resources, and operations. For example, cash-flow analytics, shareholder-value analytics, and profit/revenue analytics are increasingly important aspects of the finance function. A 2017 survey of chief marketing officers found companies spend 6.7% of their marketing budgets on analytics and expect to spend 11.1% over the next three years.16 A 2017 Deloitte survey of HR managers found over 71% of the surveyed companies see people analytics as a high priority.3 Analytics is increasingly used in operations management for demand forecasting, inventory optimization, spare parts optimization, warranty management, and predictive asset maintenance. Acknowledging extreme deficiency of data literacy among today's managers, by 2020, 80% of organizations will embark on data-literacy initiatives.  .... "

Shifting Decision Making

Quite a suggestion in the title of this piece,  Will it, can it?  Augment it in both areas enough to change the nature of decision?

Why AI Will Shift Decision Making from the C-Suite to the Front Line  by Alessandro Di Fiore in HBR

Hardly a day goes by without the announcement of an incredible new frontier in Artificial Intelligence (AI). From fintech to edtech, what was once fantastically improbable is now a commercial reality. There is no question that big data and AI will bring about important advances in the realm of management, especially as it relates to being able to make better-informed decisions. But certain types of decisions — particularly those related to strategy, innovation and marketing — will likely continue to require a human being who can take a holistic view and make a qualitative judgment based on a personal consideration of the context and facts. In fact, to date, there is no AI technology that is fully able to factor in the emotional, human, and political context needed to automate decisions. .... " 

Deep Mind and Abstract Thought

 What is abstrsact thought, and how do IQ tests track that?   Closer to what we might call common sense, or really uncommon?

DeepMind AI Takes IQ Tests to Probe Its Ability for Abstract Thought 

in New Scientist  By Jacob Aron

Google DeepMind researchers are challenging artificial intelligences (AIs) to solve abstract reasoning puzzles similar to those used in IQ tests. One particular puzzle involves looking at sets of abstract shapes and selecting which should come next in a given sequence. DeepMind's David Barrett says the researchers evaluated neural-network AIs on whether they could learn more general concepts. Standard networks performed poorly on these tests, with scores as low as 22%. However, a new neural network specifically engineered to infer relationships between different parts of a puzzle scored 63%. "While these structures help specifically with this task, we believe they can also be applied to other problems involving abstract relationships and taking decisions between possible courses of action," says DeepMind's Felix Hill. ..." 

Need for Explainable AI

FICO scores and all that.  Transparency for decision understanding.

Opening Up Black Boxes with Explainable AI   By Alex Woodie

One of the biggest challenges with deep learning is explaining to customers and regulators how the models get their answers. In many cases, we simply don’t know how the models generated their answers, even if we’re very confident in the answers themselves. However, in the age of GDPR, this black box-style of predictive computing will not suffice, which is driving a push by FICO and others to develop explainable AI.

Describing deep learning as a black box is not meant to denigrate the practice. In many instances, in fact, the black box aspect of a deep learning model isn’t a bug – it’s a feature. After, all, we’re thrilled that, when we build a convolutional neural network with hundreds of input variables and more than a thousand hidden layers (as the biggest CNNs are), it just works. We don’t exactly know how it works, but we’re grateful that it does work. If we had we been required to explicitly code a program to do the same thing as the CNN does, it likely would be a complete disaster. We simply could not build the decision-making systems we’re building today without the benefit of self-learning machines.

But as good as deep learning has gotten over the past five years, it’s still not good enough. There simply isn’t enough free goodwill floating about our current world for a hundred-billion-dollar corporation or a trillion-dollar government to tell its consumers or citizens to “trust us” when making life-changing decisions. It’s not just a wary public, but also skeptical regulators buoyed by the GDPR’s new requirements for greater transparency in data processing, that’s driving for greater clarity in how today’s AI-based systems are making the decisions they make.

One of the companies on the cutting edge of helping to make AI more explainable is FICO. The San Jose, California-based company is well-known for developing a patented credit scoring methodology (the “FICO score”) that many banks use to determine the credit risk of consumers. It also uses machine learning tech in its Decision Management Suite (DMS), which companies use to automate a range of decision-making process. ... " 

The Kinds of AI Value and Decision Problems

Nicely done view of the space and it takes it beyond JUST hard quant pattern problems, and  on to the kind of the decisions involved.  Lots of work yet to do, but this helps segment them.  Rwaad the whole piece for more detailed discussion.

The six fields where artificial intelligence (AI) will offer added value in customer experience
By Steven Van Belleghem  in Customer Think.

Six steps where AI can influence customer experience

Today (2017), we entrust all kinds of simple tasks to our virtual assistant, from setting our morning alarm call to timing how long it takes to boil an egg. Of course, this is little more than playing about. Even so, it is the start of the evolution in AI that will soon see companies offering significant added value to their customers. This will happen in six steps, each of which will result in ever-greater AI impact.

The six steps in the AI customer relationship:

1. Curation of information
2. Provision of customized information
3. Recommendations
4. Predictions
5. Automation
6. Contextual analysis   ... " 

Update: Adversarial Risk Analysis Talk

 Talk given this week by Dr David Banks of Duke University and sponsored by Yichen Qin,  Assistant Professor, Department of Operations, Business Analytics, and Information Systems,  Lindner College of Business, University of Cincinnati,  on April 6, 2018

Adversarial Risk Analysis Talk, full announcement.

Speaker: Dr. David Banks, Duke University

Title: Adversarial Risk Analysis

Abstract: Adversarial Risk Analysis (ARA) is a Bayesian alternative to classical game theory. Rooted in decision theory, one builds a model for the decision-making of one's opponent, placing subjective distributions over all unknown quantities. Then one chooses the action that maximizes expected utility. This approach aligns with some perspectives in modern behavioral economics, and enables principled analysis of novel problems, such as a multiparty auction in which there is no common knowledge and different bidders have different opinion about each other.   .... " 

Here are the slides from the talk  (Technical)

Good talk. In particular because it dealt with  how people work in interactions.   Either versus nature, or versus other humans, here in some sort of competition.  The most common game example used was the Auction.  In the cases described these were adversarial.  Under the structure of this  'game' to win the auction.   Humans in any interaction build a model of who they are interacting with.   The methods proposed construct numerical methods to define the value of alternative strategies.

 But it immediately came to  mind that these methods don't need to be adversarial.   When people converse, or ask for help from an adviser  (Human or machine).  they are looking to maximize the value of the interaction.   In a chatbot, for example a person wants to solve a problem, the chatbot has defined knowledge.  How should the interaction proceed?   How should these advisory approaches be strategically designed?  Under constraints and costs?   How do we rate Assistants in an approach to provide value?  Whats the structure of that game?    Examining.