Interaction Design for Complex Problem Solving

Developing Useful and Usable Software

Barbara Mirel

• Publisher: Morgan Kaufmann
• Year: 2003
• ISBN:
  9780080508955
• Topic: Computer Science
• Subtopic: Human-Computer Interaction
• Index: Computer Science

Software for complex problem solving can dazzle people with advanced features and alluring visuals, but when actually put to use it often disappoints and even frustrates users. This software rarely follows the user's own work methods, nor does it give people the degree of control and choice that they truly need.This book presents a groundbreaking approach to interaction design for complex problem solving applications. The author uses her vast field experience to present a new way of looking at the whole process, and treats complex problem solving software and web applications as a distinct class with its own set of usefulness demands and design criteria. This approach highlights integrated interactions rather than discrete actions, clearly defines what makes problem solving complex, and explores strategies for analyzing, modeling, and designing for exploratory inquiries.·In depth case studies ranging from IT troubleshooting to marketing analysis to risk assessments in healthcare show exactly where and what goes wrong in real world activities and how to improve them.·Presents a system and framework for analyzing complex work and takes the mystery out of eliciting patterns of work and their meanings.·Offers new perspectives for support and new design strategies for building the right models into programs so that they effectively address users' dynamic work. ·Allows designers to turn findings into useful designs for problems that require users to create new knowledge but with no one right answer and with many methods of reaching solutions.

PART I

UNDERSTANDING THE WORK

1

What Makes Complex Problem Solving Complex?

To an analyst swamped with data and struggling to stay afloat while solving complex problems, today’s computerized work world seems a pretty unfriendly place. Massive amounts of data are available for everyday problem solvers and decision makers, but the resources for making strategic use of them are paltry. For example, to figure out who is buying what and how to cross-sell to them, analysts for an e-commerce Web site must integrate and interpret a sea of data on buyers’ portal entries, click-through paths, transactions, and demographics. Similarly, hospital nurses determining dosages of medications have to work with huge amounts of medical data—often dispersed across systems including information on patients’ conditions, care plans, medication histories, allergies, laboratory tests, adverse drug interactions, prescription orders, and drug concentrations carried by the in-patient pharmacy. As the volume of data grows (and with it the expectation that problem solvers will become smarter in business and safer in operations), these problem solvers find it increasingly difficult to make decisions in indeterminate situations with confidence. From the eyes of a “drowning” analyst, with a lifetime of data passing before her eyes, problems seem to be complex mainly because of the sheer volume of information.

Enter advanced software for complex problem solving, which promises to make huge amounts of data manageable. Promoters of this software claim that with it, the people closest to the problems will get the tools they need to advantageously analyze available data and answer complex questions. Current data visualization applications for Web log analysis, for example, enable them to display and make sense of gigabytes of data in graphic presentations.

If only fixing the data volume problem was enough to facilitate complex problem solving through software support. Unfortunately, more is needed. Only part of the complexity of dynamic and emergent inquiries derives from large amounts of data. Complex problem solving is also complex because it requires reasoning in uncertainty and such higher-order analyses as cumulatively interpreting comparative relationships (sensemaking). It requires structuring your way from indeterminacy to resolution (wayfinding), integrating heterogeneous data from multiple sources, and relating multifaceted data relationships to situational conditions and fluctuating priorities (data ordeals) (Passini, 2000).

The distinguishing traits and dynamics of complex problem solving and the distinct support it requires warrant treating applications for open-ended inquiries as a separate area of software design and development. In many ways, complex inquiries are polar opposites of well-structured problems and tasks. They have vague or only broadly defined goals and inputs. They cannot be conducted through fixed rules, procedures, or algorithms. In fact, it is impossible to know all feasible solutions in advance. Moreover, unlike well-structured work, complex problem solving has no single correct answer.

Unfortunately, development groups often fail to recognize these distinctive demands and instead create applications based on assumptions and design methods used for well-structured tasks. The result is inadequate support for problem solvers’ high-level sensemaking, wayfinding, and data ordeals, and therefore, applications that are not fully useful to the people who depend on them.

Part of the reason for this failure is that development groups rarely ask the two questions that will give them insight into the thinking, methods, strategies, and choices required for designing usefulness in applications for complex problem solving. These questions are as follows:

What makes complex problem solving complex?

What support is truly useful for the distinguishing demands of complexity?

In this chapter, I address the first question to lay a foundation for the second, which is discussed in subsequent chapters in greater detail. In this chapter I provide only broad and suggestive insights into the issue of optimal support.

This chapter describes the ways in which problem solvers’ directions and possible actions in dynamic and emergent work invariably are shaped by interactive conditions within and across four main contexts: The technology and data context, work domain, problem space, and subjective context. I also identify the dynamics and traits of complex problem solving that make it complex, distinguishing it from well-structured work.

The implications are that complex problem solving requires unique support. Not only must this support address large and complicated information spaces, it also must help users configure, negotiate, and coordinate resources in landscapes of activity that are both patterned and variable. That is, software support must accommodate the structured openness of complex problem solving. Problem solvers follow certain regularities tied to professional practices, domain conventions, and business rules. However, they never fully foresee the routes and moves they will take because they organize their directions and choices around idiosyncratic situational factors, emergent insights, serendipity, and external flux.

To dramatize the challenges of complex problem solving, I start the chapter with a scenario about an ill-structured marketing problem and use it to frame the subsequent discussion about complexity and its support. The scenario reveals that a good deal of the support that problem solvers need, like Marty in the case study, is not currently built into their software. Marty’s case is the first of many in this book. Altogether, the cases reveal support for core areas of complex problem solving that users need but currently lack, by and large, in software today.

Software teams and HCI specialists do not have to design marketing applications to find relevance in Marty’s problem solving scenario. In terms of complexity, this story speaks to any team’s efforts to understand and design for users’ open-ended inquiries.

SCENARIO: SHOULD WE BREAK INTO A NEW MARKET NICHE?

For ten years, Marty has been a category manager at Quality Paper, a company that manufactures several brands of paper towels, tissues, and napkins. For years he has been solving complex problems, most of which involve figuring out how to gain greater market share for his company’s brands. Currently, he manages the multibillion-dollar category of paper towels. To see if his company is manufacturing the right goods for market demand, Marty continuously assesses the paper towel market. He grapples each quarter with new sets of data on sales, profits, promotions, competition, customer demographics, buying patterns, and best practices. He uses complicated procedures to get data into the right format and to analyze relationships. Because markets are volatile and company objectives change, no two analyses Marty conducts are exactly the same.

What Is the Problem?

For the past several months, Marty has been paying close attention to a new premium-tier of paper towels whose sales are strong. He concedes that premium is fast becoming a profitable niche in the market. His company has no paper towel product in this niche, and without one, it could lose market share. It is time to figure out whether the company should break into this part of the market.

Marty’s goals are no more concrete or specific than the previous sentence suggests. His goals will become clearer as the inquiry evolves.

Marty begins his inquiry with an hour-long meeting with his manager Nancy. They start by discussing how big a problem the lack of a premium product might be for the company. Even if results from Marty’s proposed investigation suggest it is wise for Quality Paper to enter this niche, this will likely be a tough sell to corporate decision makers. If the inquiry is to address their bosses’ greatest concerns, Nancy and Marty must turn the data he collects into persuasive communication.

The plan is for Marty to do a quick, first-run analysis at a high level and see what he discovers. He will meet with Nancy again to decide if what he finds confirms their theory. If it does, he will have to do a more thorough review and report. With that report, Nancy and Marty will create a presentation aimed at persuading decision makers to adopt the recommendation to produce a premium paper towel.

Verifying That the Problem Is a Problem

From the start, Marty’s quick, first-run analysis is slowed by the volume and structure of the data. He uses syndicated data from a national subscription service. Like other analysts, Marty uses one set of syndicated data on product and market performance and another on customer behaviors and demographics. These datasets are h...