2114 is203 - Social and Organizational Issues of Information » Week 11

WordPress database error: [Table 'i203.is203_users' doesn't exist]
SELECT * FROM is203_users WHERE ID = '1' LIMIT 1

Week 11

Apr. 3rd: Trust in Information

Marsh, Stephen and Mark R. Dibben. 2003. “The Role of Trust in Information Science and Technology.” Annual Review of Information Science and Technology 37:465-498. [PDF]

Hertzum, M., H. Andersen, V. Andersen, and C.B. Hansen. 2002. “Trust in information sources: Seeking information from people, documents, and virtual agents.” Interacting with Computers 14:575-599. [PDF]

Apr. 5th: Trust and Information Exchange

Cheshire, Coye and Karen S. Cook. 2004. “The Emergence of Trust Networks: Implications for Online Interaction.” Analyse and Kritik 26:220-240. [PDF]

Fisman, R. and T. Khanna. 1999. “Is Trust a Historical Residue? Information Flows and Trust Levels.” Journal of Economic Behavior and Organization 38:79-92.[PDF]

January 2nd, 2007
posted by:

WordPress database error: [Table 'i203.is203_users' doesn't exist]
SELECT * FROM is203_users WHERE ID = '1' LIMIT 1

13 Comments Add your own

  • 1. yliu  |  March 30th, 2007 at 10:44 pm

    Trust and trustworthiness is an interesting distinction that Marsh and Dibben raises half-way through the paper. The first is the act of having faith in the other party despite non-trivial risk; the second is the set of properties that persuade/dissuade someone from offering trust. It’s fairly important to make this distinction, especially whenever the suggestions along the lines of “designing for trust” come up. We are really only able to manipulate the properties of trustworthiness in design, and even then only to a certain extent.

    I’m not sure I can buy the idea that trust can be modeled by a process somewhat akin to adding up all the “trustworthy” properties (things that inspire trust) and weighting it by some subjective constant, subtracting all the properties that incite distrust, and checking against a personal truth or threshold of sufficient trustworthiness, but as least it’s a computable model. After all, even human perception of trust is more or less heuristically oriented.

    The “wearable” trust/reputation machine was a quick aside that the authors tossed in, but seems a fairly interesting ideal concept. Our personal experiences is a pool of information that we use first to assess trustworthiness, but is also a fairly sparse set of data. We often have to be fall back to such things as first impressions/surface observations and general stereotypes (Hertzum et al). If there existed an ubicomp device that would store your own reputation as rated by other users, and allow you to rate other users on the fly as you go about daily business, then that neatly solves the problem. It’s a very different model than having either a massive centralized reputation-vouching authority, or a “chain-of-recommendations” based system that we tend to use (”I trust Foo, and Foo trusts Bar, and Bar says that I can trust Baz”). In practice, this might not work out so well. The data sparsity problem is simply made recursive (what if there isn’t enough data about this particular person? how do I know if I can trust the judgment of these other users who rated this fellow), but as a more “P2P” alternative solution to the trust problem, it bears some thought.

    The COGITO experiment with virtual avatars was interesting and informative. I dunno, I think it’s premature to conclude that people just simply find human avatars more trustworthy machine avatars, though. Humans are just used to dealing with other humans. Since trust is all about experience and heuristic application of experience, these people would of course tend to prefer human-represented avatars. This is doubly the case with the well-known actor, considering that this actor probably has a prior and known reputation, and that is factored into the word association assessments (even though it’s just a representative avatar). Being around animated avatars from various computer games and movies and such, I honestly don’t think I’d rate or trust any of the human representations any more than the computer-generated models. There are always the weird, almost-human, “uncanny-valley” kind of avatars that give people creeps, but those are of a different cause entirely.

  • 2. daniela  |  March 31st, 2007 at 10:17 am

    I enjoyed both Marsh and Hertzum’s discussion of whether trust can only be associated with humans, not inanimate objects or entities without agency. Hertzum et al. initially puts forward the idea that “trust is inherent in all human relationships” yet their research and others have find that design of social cues can allow for trust relationships with non-human agents. Marsh’s critique of this work rings true for me - that most of these studies didn’t properly take into account the culture in which the trusting agent is situated. In this sense, culture could influence the ability of the information seeker to correctly interpret the communicated information as well as change the meaning of ‘correct’ interpretation in the given context. Hertzum et al’s studies of virtual avatars illustrate how important culturally determined factors may be in the assessment of these visuals. People’s reactions to recognizable or likable actors would depend on the society or culture to which they belong.

    I felt Marsh was mistaken in focusing on certificate authorities (CA) as the trusting entity, since the technological literacy of each player weighs heavily in being able to understand or even recognize the meaning of a CA. But, as he points out, social factors and public opinions are likely far more instrumental in assessing the trustworthiness of a system than the technology itself (much like assessing the efficacy of politicians).

    Marsh mainly highlights the idea that an entity acts as a representative of the group or set in which a trusting party places trust. I agree with this implication that interpretation and subjective assessment of a representation are key components of one’s measurement of trustworthiness. His analysis ties into my interest in the role that graphic structure and aesthetics in subjective emotional judgment, exemplifying the relationship-based dynamic model of trust. Also interesting is the type of link Marsh brings up between high trust societies, and economically strong societies, and the looseness of network ties supported by the web. I wonder whether an adaptive relationship could exist between group/individual/object and the assessment of trustworthiness by individual in a form that is distinct from the assessment by society as a whole. Do high trusting societies form different value judgments from low trusting societies based on an adaptive process, such as the interpretation of aesthetics? It seems this relationship could be either casual or correlated, but I’d be curious to see if aesthetic decisions correlate with the economic strength of that society.

  • 3. Bernt Wahl  |  April 3rd, 2007 at 9:26 am

    Trust is important, children need to trust their parents for guidance, a patient needs trust their doctor and people learn to trust based on signs.

    We look at telltale signs to see if we can trust individuals based on past experiences context at a bar vs. at a church.
    Trust the information from a reportable source a trusted news paper a science group or trusted organization
    Trust in ecommerce is the information coming from a reliable site, could con-artist spoof the site to gain your trust.

    In 1996 we built an eCommerce a commence site for Microsoft, Sun and Macromedia. SSL still had not been incorporated so to satisfy our supplier clients, we created a credit card encryption system. They remained uncomfortable with our system until I create a fancy seal that said encrypted by “… system” and no one mentioned anything after that.

    Items that build trust in a web site design are.
    • Clean professional look
    • Accurate information
    • Pictures of professional
    • The site works, no dead links
    • Clear since of purpose for the site
    • Links from other professional sites
    • Good use of color; Black, white and blues
    • Correct grammar usage and spelling

  • 4. cvolz  |  April 5th, 2007 at 4:08 pm

    Trust seems to pervade a lot of human interaction. While reading these papers it occurred to me (partly just to be contrary) to try to envision what a completely untrusting populace would be like and it seems that some element of trust between indviduals is required for almost any sort of cooperative action.

    It also occurred to me that because trust is such an integral part of human interaction that people may be biased to trust rather than against. And, given the relatively arbitrary attributes and qualities that people look to, absent any other more quantitative measures of trust, people seem to almost be looking for a reason to trust someone. Just because they “look” trustworthy.

    I do wonder, however, if the “look” of being trustworthy is based on or enforced by stereotypes or formed by earlier experience of other people who turned out to be trustworthy.

    It would be interesting to see if children prove to be naturally trusting or not trusting or trust-neutral. Also, it would be excellent excuse to perform more experiments on children.

  • 5. evynn  |  April 6th, 2007 at 1:16 pm

    There is an element of reciprocation related to risk and uncertainty* that I find fascinating, and is definitely worth addressing more fully. For the most part we’ve talked about how one person forms trust in another, but human interactions cannot be reduced to one party trusting and one party being trusted. There is always a (usually silent) negotiation of risks on the part of both parties, and usually the risks and uncertainty are highly uneven. Taking the airport example of asking someone to watch your bags, you try to find someone who appears trustworthy based on some conscious or unconscious criteria, but once you ask, that person must also assess whether they trust you not to abandon them with a bag full of drugs or a bomb they’ve gotten past security. So the risk for the person leaving the bags is small (the person might walk off with them) but the risk to the person watching the bags is potentially large– they might be endangering their life or freedom for a complete stranger. Perhaps the watcher will ask how long the leaver will be gone or which bathroom they’re using and plan to call security if they’re not back in a reasonable time.

    There is of course a reason that people agree to watch other peoples’ bags in airports though, and it has to do with the uncertainty factor: The person leaving the bags may not be totally certain that their chosen watcher isn’t a petty thief, but the watcher is usually much more certain that the leaver is not involved in an elaborate scheme to frame or kill them. Even in the age of the “War on Terror” people have a general understanding that such things are improbable and furthermore, a high enough degree of trust in airport security to believe the screening process has made them even less probable.

    It would be interesting to think about power is defined in these unequal risk-uncertainty relationships. We touched on this when we talked about trust without choice- if we trust someone because we have no other choice, is it meaningful to talk about uncertainty and risk? We may not think trust in the person has much chance at all of delivering a beneficial outcome, or we may be highly uncertain of them. I doubt there are many situations in which there are truly no alternate choices, but if the alternatives involve a high certainty of a very negative outcome, then the person we’d prefer not to trust because of uncertainty or risk that is equal-to-or-less-than the alternative suddenly becomes a very powerful person. It seems that it’s more meaningful to talk about trust when the parties have unequal levels of uncertainty and risk, and furthermore the levels possessed by each party are complementary (eg, high uncertainty but low risk for one, and low uncertainty but high risk for the other).

    *I’m referring to risk in the sense of possible negative or positive outcomes associated with trust, not the chances that they will occur- that is what I’m referring to as uncertainty.

    238a
  • 6. srini  |  April 7th, 2007 at 1:10 am

    It is interesting to analyze the notion of trust in information sources, as done by Hertzum et al. In the beginning, the authors described the notion of trust and later on analyze two different cases. I would agree with the author’s statement that trust is inherent in all human relationship. It has been stated that trust is at play when people exchange information, but I feel that people exchange information only when they trust the other party, either it is a human or an agent. And moreover trust is built over a period of time and not instantaneously. Hence the factors such as relevance and usability may help in building the trust.

    Among the factors used for the study the ‘people or documents’ and ‘information or commitment’ seems more interesting. We can see a connection between these two factors. Documents are likely to carry specific information but not commitment, but people can show commitment. Though it is absolutely from my point of view, I based my argument based on the fact that there is no hidden content in a document and hence it can be read and the information can be understood. But the same information presented in a document can be perceived in different ways by different people. I feel that factors such as appropriateness and up-to-dateness are important but interpretation is more important than the other two. A committed human being can handle all the factors mentioned above in the case of documents, but he cannot be read explicitly. Hence the notion of trust plays a major role when the information source is a human.

    The implications of COGITO case, the study of trust in virtual agents is relevant to current e-commerce generation. Though a surface evaluation of the agents appearance may not imply the trustworthiness, humans would feel better when the agents have a human face.

    As Internet is becoming one of our major sources of information, agents are becoming popular over the internet to promote e-commerce and to make extensive use of customer data. I could see the comparison of trustworthiness between agents and people as a comparison between automated recommendation systems and social network recommendations like yelp. Though automated recommendation seems to work in case of book recommendation like amazon, it may fail in cases like restaurant recommendations. A further study in this area would reveal interesting results. Trust becomes a major factor when the agents request for the customers personal data such as credit card, SSN, etc. I feel that in such context people trust the agent mainly based on the vendor and not based on the appearance of the agent. Other factors such as reputation and recommendation (by others) also play an important role in developing a trust. As concluded in the paper, I would agree that the strengths of people, documents and agents vary and they can be combined effectively.

  • 7. karenhsu  |  April 8th, 2007 at 1:31 am

    The COGITO experiment was an interesting study on user requirements of personified virtual agents in e-commerce. I agree with Yiming’s assessment that at present, humans are simply just not used to dealing with animated avatars, which might be behind their being perceived as less trustworthy than human-represented avatars. However, also having been plenty exposed to animated avatars from video games and films, I’d say this familiarity produced the opposite reaction from me. The focus groups from this experiment list seriousness as a strategic requirement for virtual agents to contribute to effective e-commerce, something that I agree with. However, the contexts of both video games and films usually produce a less than serious state of mind for me.

  • 8. elisa  |  April 8th, 2007 at 9:10 pm

    I recently read an article on Science Magazine about Mental Health in Developing Countries. It said that “many epidemiologists have reported low rates of depression and other mood disorders in East Asia, including in China. A survey coordinated by the World Health Organization (WHO) found that roughly one in 50 people in Shanghai and Beijing suffered from a diagnosable mood disorder over a recent 12-month period. In the United States, one in 10 had, according to the survey, published in the 2 June 2004 issue of the Journal of the American Medical Association. However, surveys of this sort have a flaw that may skew results: they are designed to detect disorders as experienced by Westerners, says Arthur Kleinman, a medical anthropologist at Harvard Medical School in Boston.” The fact that these researches are usually designed by Western (or Western-educated) scholars and might therefore reflect the Western cultural tradition is an issue that has constantly been in my mind while doing many of this class readings (and I use the world ‘Western’ with a certain reluctance, given that I often find the American viewpoint/starting point quite far from the European one). My perplexity flared up in reading the Fisman-Khanna article: is ‘trust’ a concept like ‘depression,’ that is something that has been encoded in Western research, and indicates something that exists indeed in every society, but is not quite expressed and, more important, conceptualized, in the same way? Sometimes, to find out attitudes towards a specific issue, different questions need to be asked in different societies (take for example something like privacy: the word doesn’t translate directly into Italian – the literal translation is secrecy, which is something different from privacy; the concept exists, but it has developed in a very different way from the Anglo-saxon world, so much so that recent laws on privacy use the English word “privacy”). Fisman-Khanna, or even the World Values Survey committee whose data they use, do not seem to consider this aspect at all (take a look at the photo of the executive committee of the World Values Survey to see what I mean). The Inglehart-Welzel Cultural Map of the World they use looks very suspicious, with its very granular view of European cultures and very broad, 10,000 feet views of everything else. The core of Europe is placed in all these differently nuanced positions, but Africa is all in the survival/traditional values area? How is this constructed, exactly (I read the explanation, but where do the data come from?). The six basic cultures that Fisman-Khanna borrow from Huntington didn’t quite work even in his erudite and scholarly book, with quite a few countries constrained in roles that were not theirs so that they could fit the societal (which was actually religious) divisions; here, without the background of knowledge and analysis that at least Huntington provided, it looks superficial and facile. The jump to mapping trust to technology is even more ludicrous (especially after watching The Lives of Others)…

    2843
  • 9. nfultz  |  April 8th, 2007 at 11:48 pm

    Trust is a pretty cool topic, AI people have done some stuff on it with modal logic:

    A modal logic framework for multi-agent belief fusion

    Mutual enrichment through nested belief change

    Also, a neat paper on datamining ePinions for trust networks (with negative trust):

    Propagation of trust and distrust

  • 10. bindiya  |  April 9th, 2007 at 1:00 am

    It was interesting to read the paper on the emergence of trust networks under uncertainty. I am particularly intrigued by how reputation systems work, since I am working on designing one for the User Interface Design Class. Most online interactions are devoid of many of the cues that people use in the physical world to make judgments about the character, stability and reliability of people and systems. Yet these cues are critical to sophisticated interactions and transactions and online reputation systems are an attempt to cover this gap. The issues we are dealing with while designing an online reputation system is considering the impact of positive vs. negative vs. mixed reputation systems. In a relatively closed system people are less likely to change their real name and a positive reputation system might ultimately produce better quality work. The weakness and disadvantage of reputation systems is that there is always a potential for attack by people wanting to game the system, whether to gain attention for themselves or their point of view or to defraud other users. This is an area that needs to be paid attention to. Also it is important to realize the distinction between trust and reputation. Trust can be defined as expectation of people towards an entity which affect their choice when an action must be taken. Reputation on the other hand is a basis for trust.

  • 11. jilblu  |  April 9th, 2007 at 6:27 am

    Regarding Fisman and Khanna’s piece on trust:
    It’s certainly intuitive that trust varies across different cultures, but Fisman and Khanna’s results seem to contradict some prevalent stereotypes. For example, people from “primitive” cultures are often described as more naïve, more trusting. Yet the study’s results find a correlation between urbanism and trust. As explained in the study, network densities in urban communities may be too low to provide the deterrence-based trust present in smaller communities. Yet the study shows no correlation between community size and trust.

    The study measures two-way communication by counting number of phones per capita; it doesn’t measure network density at all. If it did, perhaps there would be a correlation between network density and trust, which may be more consistent with common stereotypes of trust in urban verses “primitive” cultures.

    I’m really surprised that Fisman and Khanna chose to measure trust with just one question: “Generally speaking, would you say that most people can be trusted or that you can’t be too careful when dealing with people?” Since they were surveying thousands of people across different countries, I can’t imagine why they didn’t ask a range of questions regarding trust, and then give the respondent an overall score based on their answers.

  • 12. Ken-ichi  |  April 9th, 2007 at 5:58 pm

    On Marsh & Dibben

    I like the idea of the “locus of trust” when assessing how people think about trusting organizations. I guess I’m of the opinion that people are generally willing to trust an organization as a whole when that trust is base dupon past experience. I won’t buy an Apple laptop because I personally trust Steve Jobs, or Apple engineers, or Apple UI designers, or Apple marketers. I’ll buy it because I’ve been happy with my current one. One the other hand, when there isn’t past experience to go on, I suspect people are more wary of organizations and begin to shift their loci to smaller subunits. Apple has never made a phone before, so should I trust the organization as a whole and buy the iPhone when it comes out? Perhaps I should scrutinize Apple’s hardware engineers and their ability to make reliable portable devices (pretty good), or their UI designers for their ability to make usable interfaces with small screen real estate (arguable), or perhaps I should examine Apple’s marketing department for their ability to market the product well enough that the iPhone will be popular enough that I never have to worry about incompatibility with third party products and services, and I never have to worry about being saddled with dead-end technology (can anyone say “Newton”?).

  • 13. jess  |  April 10th, 2007 at 7:34 am

    Marsh and Dibben make the point that those in positions of power and authority in an organization represent the trust of those within an organization (473). I experienced this daily when traveling in Peru and Bolivia two years ago. It was really interesting, and not at all surprising, that when I said was from the United States people did not seem to trust me until we discussed politics and I revealed who I voted for. Once this was cleared up, trust was not an issue. For this reason, I met a lot of Americans who introduced themselves as “I voted for Kerry.” But these issues in trust came as no surprise; the politics and policies of the president are widely displayed in the international news.

    This just goes to show that trust depends on the context. While I really enjoyed reading the COGITO virtual agent experiment, I couldn’t help but think that the method did not consider context of trust sufficiently. Different virtual agents will be trusted differently depending on their context of use. For example, I associate the Lego man with fun, building, and playing. Therefore, I would be more likely to trust this agent in the context of an online toy store or other stores that are considered fun and playful. However, I would not trust this agent to give me advice on choosing hair products or helping me find a job. Another example is the actor virtual agent, whom I often associate with beauty and fashion. While I would not trust the Lego character to help me choose hair products, I would be more likely to trust the actor in this situation because she has nice hair (and not a plastic wig.) However, I would not trust the actor to help me buy a blender or other cooking items. Therefore, while the experiment was correct in making the general inference that a scary looking virtual agent is not as trustworthy as a kind and gentle agent, it depends entirely on the context; the very same scary agent might be deemed more trustworthy at a Halloween costume site.

    7eb

Leave a Comment

You must be logged in to post a comment.

Trackback this post  |  Subscribe to the comments via RSS Feed


0