Who Are the “End Users”? Mary Shaw Carnegie Mellon University

There are lots of EUs C. Scaffidi, M. Shaw, and B. Myers. Estimating the Numbers of End Users and End User Programmers. VL/HCC'05: Proc 2005 IEEE Symposium on Visual Languages and Human-Centric Computing, pp , Using data from the Bureau of Labor Statistics, we estimate that over 90M Americans will use computers at work in Of these, only about 2.5M will be professional programmers; 40.5M will be managers and (non-software) professionals. This does not include home users or non-US users, so there will be many more than 90M total end users. Most of them will “program” in some way.

They are not all alike C. Scaffidi, Andrew Ko, B. Myers, and M. Shaw. Dimensions Characterizing Programming Feature Usage by Information Workers. VL/HCC'06: Proc2006 IEEE Symposium on Visual Languages and Human-Centric Computing, pp , Analysis of web-based survey of Information Week readers

Their skills differ, even within clusters  “Programming” can mean anything from editing spam filters to writing complex code Copying the html for a web page hit counter vs creating that html yourself  Let’s not make a sharp distinction about where “programming” starts  Let’s not lump all programming together No citation yet,

Therefore we should …  Identify some structure for the vast population of EUs Find more content clusters Create something like Bloom’s taxonomy for skill levels  Then we can profile tasks or EU groups “level 4 in macro cluster, level 2 in link cluster, …”  Then we can focus on the needs of groups with specific profiles

EUs are not SEs  EUs do not have rich and robust mental models of their computing systems they fail to do backups They misunderstand storage models (especially local vs network storage) they execute malware they innocently engage in other risky behavior.  The responses of SE to this mismatch between real computing systems and EUs’ models has been to seek ways to help the users act “rationally”.

Real EUs’ decisions are not rational  Systematic errors in frequency estimation People systematically overestimate low frequency events, overestimate high frequency events  Which is more common, murder or suicide? Memory effects: easiest retrieval is taken as fact  Recently viewed information (news) dominates collected facts (data)  Subjective judgments Peoples’ perceptions of incidence and risk are notably (and demonstrably) flawed  Simple models Simple linear functions consistently outperform experts Insensitive to coefficient values (sensitive to sign) Reid Hastie and Robyn M Dawes. Rational Choice in an Uncertain World. Sage Publications 2001 Reviewed in Reid Hastie and Robyn M Dawes. Rational Choice in an Uncertain World. Sage Publications 2001

Therefore we should …  Stop trying to “fix” the users, but instead support the ways they actually think Metainformation structures that accumulate positive and negative evidence  Develop EU SE support that supports the forms of reasoning that real people use “low-ceremony” evidence (reputation, reviews) as well as “high-ceremony” evidence (verification, testing)

End of presentation

High Ceremony Evidence  Widely accepted among computer scientists  Potentially high levels of assurance  Need precise specifications, substantial effort  The Academic Big Four – the “gold standard” Formal verification Results from trusted automatic generator Systematic testing Empirical studies in operation  And also Inspections Assurance cases, other sound certification (others from comparative analysis)

Low Ceremony Evidence  Widely available information, used informally  Largely ignored by professionals  Not suitable for high assurance, but inexpensive  Examples “best X” reports (linear functions of subjective marks) editorial reviews recommending certain components for certain contexts (cf Consumer Reports) advertising claims by vendors 3rd party reviews of vendors and products by users recommendations by co-workers auction and betting mechanisms, “wisdom of crowds”, branding, seller reputation subjective certification checklists popularity