agents:agenda:a definition:origins:origins (continued)J. Lanier: “My Problem with Agents”Lieberman: “Autonomous Interface Agents”Slide 8Slide 9Slide 10Slide 11J. Seely Brown; P. Duguid: “Agents and Angels”PowerPoint PresentationSlide 14Slide 15Slide 16Slide 17Slide 18D. Norman: “How Might People Interact …”Slide 20Slide 21Slide 22P. Maes: “Agents that Reduce Work…”Slide 24IntelliShopper (University of Iowa – 2002)Slide 26“Pattie” (on Pattie Maes)bots & agents todaySlide 29KQML - Tim Finin (1997), UMBCKQML - StructureSlide 32Agents – new researchAgents – further reading01/15/19 i385t KMS 1agents:amanda mccoy, dept. of design01/15/19 i385t KMS 2agenda:General background.Highlights from readings + questions.Current research, possible futures.01/15/19 i385t KMS 3a definition:Agents have several definitions. In a nutshell they are any entity that has been programmed or has “learned” to autonomously act either directly, or on behalf of a user to perform a task.Agents can facilitate, learn, respond, initiate, react, anticipate, and behave… according to instructions or rules. This is, however highly controversial.01/15/19 i385t KMS 4origins:Software agents evolved from hardware agents, or control devices dating back to World War II.Robots are, conceptually, the hardware version of agents, though we tend to think of them more in the realm of artificial intelligence. Agent research is a sibling to A.I. It makes sense to link an application programmed to respond based on human rules of behavior (A.I), to a program we directly command to act on our behalf (agents).01/15/19 i385t KMS 5origins (continued)The term “agents” was coined at MIT in the ’50s by John McCarthy & Oliver Selfridge. Several companies and research labs jumped on the idea in the 80s. (Apple, MIT…)By the 90’s agents were implemented in many everyday applications such as email.Most agent technology these earlier articles discuss seem to be invisible today, or has become part of what we consider the application…Thoughts?01/15/19 i385t KMS 6J. Lanier: “My Problem with Agents”Lanier opposes putting too much emphasis on letting agents have autonomy, and leading users to feel their agents are “smart.”One must ignore the “guts” of an agent for it to seem autonomous. Once you look at the guts, it’s direct manipulation, just as we had before.“Bad science thrives on the most difficult problems…” Is this bad science?01/15/19 i385t KMS 7Lieberman: “Autonomous Interface Agents”Lieberman states two branches of Agents:•Interface agents-oActively assist users in an interface.•Autonomous agents-oTakes action without intervention and operates concurrently.Interface agents (examples)•Context sensitive help systems•Intelligent tutoring systemsInterface agents provide a “way-out” of having to design interfaces using a one-feature-per-task method.01/15/19 i385t KMS 8Lieberman: “Autonomous Interface Agents”Autonomous agents•Are always running•Can operate in parallel while user completes other tasks.•Can operate behind-the-scenes.•Are the exact opposite of a command line interface.01/15/19 i385t KMS 9Lieberman: “Autonomous Interface Agents”Autonomous Interface Agents•Have a front-end interface, but complete the task autonomously.•Example: You receive a notification in email about a update to a web page. •Doesn’t it seem odd to call these agents, when we often refer to them as “preferences”? Consider web-pages sending you updates to your phone or email? We simply select this option in a preferences web page. •Are web services agents?01/15/19 i385t KMS 10Lieberman: “Autonomous Interface Agents”Letizia is an autonomous agent for browsing the web. •…channel surfs by keeping one or more window open and shows pages that might be of interest.•…makes “bolder” recommendations instead of accurate judgments. •Makes use of breadth over depth to recommend pages nearby.Letizia notices serendipitous connections •presents them to users (similar to browsing books on the same shelf in a library).Works best in situations where decisions are not critical. •This is a common thread in conversations concerning agents. We don’t want to leave a do-or-die task to an agent. Do you?01/15/19 i385t KMS 11Lieberman: “Autonomous Interface Agents”Imagine if the “Google Viewer” worked more as an autonomous agent modeled after Letizia? We would probably get much more mileage out of it. Example: Users searches for “knowledge management systems” in Google and based on her past browsing preferences, Google stops and “pings” her when it finds one it thinks is a “good guess”. Would you pay for something like this now?01/15/19 i385t KMS 12J. Seely Brown; P. Duguid: “Agents and Angels”“Chatterbots” simulate human response. Include “Eliza” and “Shallow Red” (two infamous bots that fooled people into thinking they were human). They are autonomous agents. Consider Airline phone reservation systems. AA allows users to actually say “agent” to speak with a reservation agent. Search engines are the underpinnings of most agent software today. Most internet retail sites using collaborative filtering are examples. They seem to be considered a big part of agent technology in these articles.01/15/19 i385t KMS 1301/15/19 i385t KMS 14J. Seely Brown; P. Duguid: “Agents and Angels”Other examples include Sherlock on the Mac, which aggregates multiple web searches via the desktop UI. NASA has a smart t-shirt that can detect vital signs and can summon advice as necessary. Broker agents alert users when items of interest are on sale etc. You get the picture. 01/15/19 i385t KMS 15J. Seely Brown; P. Duguid: “Agents and Angels”Trust is a major issue where agents are concerned. As users discover reasons to find agents untrustworthy, their confidence in using them plummets. Nobody wants an agent to try to sell them anything, but OOPS they already do!Specific examples from Amazon.com and the Sabre system of flight recommendations have both been examples of this. Who controls the agent?01/15/19 i385t KMS 16J. Seely Brown; P. Duguid: “Agents and Angels”Negotiation is something agents don’t do well. Humans negotiate using complex methods. “…usually humans negotiate behavior more often than price”.Negotiation is a dynamic process, and humans misbehave, so why should we expect agents not to? We keep striving to