• Archive

Why Your E-mail Requests Are Ignored

Sending e-mail requests for help or information to multiple people will get you an answer more quickly, right? Wrong. Greg Barron on the importance of personalized e-mail.

We've all done it. We need a piece of information quickly, or we want to get immediate attention from IT support. So we send off an e-mail to multiple recipients—the more lures in the water, the faster we catch a fish, right? Well, not exactly, says Greg Barron.

Barron, the CLER Research Fellow at Harvard Business School, and research partner Eldad Yechiam, a post-doctoral research fellow at Indiana University's Department of Psychology, have studied why online help requests are sometimes ignored. Barron discusses the research and its practical applications in this Q&A.

Wendy Guild Swearingen: Your research into online and e-mail help requests and responses is fascinating. How did you become interested in the topic?

Greg Barron: Like our first paper's fictitious hero, Sarah Feldman, I wanted responses to my e-mail. I was looking for an address and I knew of two people I could ask. The question was, how many e-mails (asking for the address) to send, one or two? One e-mail addressed to both people is obviously the most efficient option in terms of both my time and the net's bandwidth. However, my intuition, and that of my co-author Eldad Yechiam, was that two individually addressed e-mails would be more effective. Our reasoning was that one e-mail addressed to both recipients could lead to a diffusion of responsibility where each recipient assumes that the other will respond. We couldn't resist testing the hypothesis empirically and so these experiments were born.

Q: It seems counterintuitive that the more people that are queried for help, the fewer respond. Were you surprised by your results?

A: We demonstrated that the more people queried, the lower the proportion of responses. While we were pleased with the clean results, we were not surprised. Social psychologists have been studying the diffusion of responsibility effect ever since Darley and Latané's (1970) influential studies that were motivated in part by the murder of Kitty Genovese in full view of thirty-eight bystanders who did nothing to help. It seemed natural for us to assume that the effect could be generalized to e-mail requests.

Q: Can you explain a little bit about social cueing theory and how it applies to your research?

A: Latané and Darley explained their findings in terms of the bystander's cue value—the belief, conveyed by verbal or nonverbal communication, that others are capable of helping. Accordingly, if an e-mail sent through a discussion group is evaluated by its recipient as being sent to many individuals that are capable of responding, the diffusion of responsibility effect would imply a decreased tendency to respond.

Q: Do you think there is a way to "unlearn" the diffusion of responsibility and, if so, could the change be implemented in service-based industries? On that note, have you found that there is a certain personality type that is more prone to answer online help requests sent to multiple e-mail addresses?

A: Economists model diffusion of responsibility as a ''volunteer dilemma'' where the probability that a rational person will volunteer to produce a public good decreases with group size. The fundamental part of this dilemma is that the utility of not volunteering is higher than the utility of volunteering, assuming that someone else has volunteered. The key to making the dilemma (and the diffusion of responsibility) disappear lies in increasing either the cost of not volunteering (i.e., of shirking) or the personal gain from choosing to respond. The easiest way to do this is simply to designate responsibility. In this context it is interesting to note that responsibility literally means the ability to respond.

Managers need to keep their e-mails personalized whenever possible. It's that simple.
— Greg Barron

While we have not looked at specific personality types, we did find that requests sent under the female name, Judy Lamson, had a slightly, but significantly higher response rate than requests sent under the male guise, George Lamson. This finding is consistent with Eagly and Crowley's (1986) meta-analysis on the effect of gender on helping behavior. Specifically, they found that people tend to help women more than men. Our study cannot conclusively support this result since we examined only two senders' names.

Q: In my own experience, when I have a computer-related problem I often "cc" many people in a group assuming that the more people I include, the quicker the response will be. Obviously, this is a mistake. How can groups, such as IT help desks, insure that help requests have been answered?

A: The effect of additional addresses in the cc field on the recipient is an interesting empirical question that we have not looked at, but note that in terms of cue value, the cc field is very different from the To field. Almost by definition, recipients in the cc field are not expected to respond to the e-mail so we would not expect a diffusion of responsibility to occur. As for customer service operations, it's all about designated responsibility. It must be crystal clear who has the responsibility to respond to a call, no matter how many copies of the call go out. This is a good example of increasing the cost of not volunteering, but for the one person whose job it is to respond. The person who has the responsibility cannot afford not to respond. Clearly, the cost here is in the context of professionalism and a good work ethic.

Q: How can business managers incorporate your findings into their day-to-day operations?

A: Managers need to keep their e-mails personalized whenever possible. It's that simple. The idea that a personalized communication has a larger impact is supported by a large body of both psychology and marketing literature besides our own line of research. While this all sounds intuitive, I never cease receiving e-mails addressed to undisclosed lists. The fact that some of these e-mails are actual commercial offers suggests that potential profits are being lost by not following this intuitive guideline.

There is a bit of a technological barrier here. Most e-mail clients are simply not equipped to send out personalized e-mails based on a list. This is not surprising since the need here, for the personal touch, is not technological but psychological. Paradoxically, we need e-mail clients that can do the job less efficiently from a technological perspective.

Q: What are you working on now?

A: Broadly speaking, I am interested in both decision-making and strategic interaction and in their application to managerial contexts. As for this particular line of research, I am looking at implicit signals of an e-mail's value. Value in this context might refer to the quality or importance of information in the e-mail or to the desirability of an e-mail offer.

In biology, and more recently in economics, the theory of honest signaling provides some possible insights. According to the theory, signals that are costly to produce and send provide a mechanism by which two parties can have reliable communication despite conflicting interests (Zahavi's handicap principle). Examples include male peacocks that use costly ornaments to display quality to potential mates and baby birds that use costly begging calls to display hunger. Eldad Yechiam and I think that there are analogous costly signals in e-mails that signal their quality and, as a result, increase our tendency to read and respond to their message. In the near future we will be running an experiment that examines this hypothesis empirically.

Wendy Guild Swearingen is the publications coordinator for HBS Working Knowledge.

Greg Barron is the CLER (Computer Lab for Experimental Research) Research Fellow in Business Administration at Harvard Business School and at Research Computing Services.

Learning to Ignore Online Help Request

by Eldad Yechiam and Greg Barron

The present study examined the effect of shared responsibility on responsiveness to Internet help requests. The results are consistent with recent findings demonstrating that diffusion of responsibility effects are not limited to the physical world but can also exist in a virtual world where the presence of others is indicated by the e-mails they generate (see Barron and Yechiam, 2002). In line with the assumed role of cueing, recipients of personally addressed e-mails were almost three times as likely to comply than were recipients who received the e-mail from a discussion group.

The present results indicate that even without an explicit knowledge about the presence of others, a diffusion of responsibility can occur. In the discussion group e-mail there is no explicit account of the number of individuals sharing the request. This implies that signals that an e-mail is sent to multiple individuals (such as an e-mail sent to "undisclosed recipients") have an important role in eliciting diffusion of responsibility.

Additionally, the current results suggest that the ignorance of help requests may be the outcome of a learning process. A large negative correlation was found between the number of subscribers in the list and responsiveness to the help request. This pattern appears to correspond to the behavior in simple shared responsibility games (e.g., Erev et al., 1995), where the player does not know if others share the responsibility, but learns that pro-social behavior is not reinforced.1

The alternative explanation, that participants of large groups get more help requests, and as a result, have fewer resources available to help, also had merit. Yet, the correlation between the amount of traffic on a listserv and the tendency to help was relatively small. The more robust result is that the number of valid members of the listserv effected the propensity of helping. We believe that since this result is based on a non-randomized variable it should be treated cautiously. However, it is consistent with the findings of Erev et al. (1995) and Gopher et al. (1995), who demonstrated that in a shared task, when the presence of others is not known, the diffusion of responsibility effect is produced through many repetitions of the task.

In real life the two factors studied here, the effect of cueing and of experience, often coincide. In large discussion groups they may create a "disobliging society." First, social cues, indicating that a large number of people have received the same e-mail request, are evident in the server signature on any e-mail sent to the list. Secondly, individuals adapt to a situation where the responsibility for answering help requests is mostly shared. It appears that through technology individuals learn about the responses of others to help requests (as opposed, for example, to postal requests), but this learning may not always increase the tendency behave pro-socially.

The present results are consistent with other findings related to diffusion of responsibility in large e-mail communities. For instance, in virtual discussion groups the "task" is the contribution of one's thoughts and feelings and the outcome is sharing others' experience. Therefore, individuals who do not generate new postings and only enjoy others' experience without contributing their own are social loafers to the effect that if all individuals behave this way, the community suffers. Smith (1992), for example, found that 1 percent of the 7000-person user-population generated 50 percent of all postings to a commercial Bulletin Board System. The remaining 99 percent merely "lurked." Likewise, Rojo and Ragsdale (1997) found that 82 percent of the user population of academic e-mail forums never contributed, but only read the dialogue of others (see also Zhang and Storck, 2001).

One situation in which the diffusion of responsibility phenomenon within e-mail communities may not occur is the case where the response to the request for help is not only of private interest but of public interest as well. For example, a person who seeks advice about a topic that interests other list members (e.g., a certain game-move in a chess related group). In this case the answer may be sent to the entire virtual community, rather than just to the requesting individual, and be rewarded by members' responses (such as approval or admiration). In line with social facilitation theory (Zajonc, 1965), for these types of requests the presence of active virtual bystanders has been found to enhance responsiveness (see Rafaeli and Noy, 2000).

Taken together, these findings should be considered by designers and users of discussion groups, as well as marketers who seek to use discussion groups to advertise their products. Users of discussion groups who seek help that they wish to be delivered to them individually (being of no interest to the general public) may benefit from sending e-mails to these participants individually.

Designers of discussion groups may seek to create modes of interaction that will benefit those who seek help. For instance, they might create a sub-society consisting of individuals with a high response rate. Its own members could address this society separately from the larger discussion group. When the expression of opinions is the goal of the e-mail, individuals will possibly want to share their opinions with the larger group. On the other hand, in the case of a help request, it would be better to send the e-mail separately to the smaller group. An indication—on the e-mail—that it has been sent to the smaller group might serve as a positive cue for helping.

A similar system is used in mIRC chat channels, where users can be given a special signal (a "+" sign next to their name on the list) indicating that they are special members. These users are not the channel administrators, but members chosen by the administrators as being extra-contributive to the channel. The special status is mostly symbolic, with little added benefits. On the other hand, it possibly serves as an important social cue about the "+" members. The effect of mIRC's hierarchical organization on the diffusion of responsibility and other cooperation dilemmas is a stimulating topic for further study.

Finally, an important application of the present results is related to marketing strategy. Discussion groups are often the targets of messages designed to persuade the recipients to go to a certain Web site. While there is a clear incentive to send such messages to discussion groups, since a single such message reaches thousands of recipients, the current results suggest that there may be a price to pay for this strategy in terms of the recipient's response rate.

Excerpted with permission of the authors from "Learning to Ignore Online Help Requests," by Eldad Yechiam and Greg Barron. Forthcoming in Computational and Mathematical Organization Theory.


1. Furthermore, members of large discussion groups were less likely to respond to a help request even when the request was individually addressed. This supports our assumption of a potential generalization from the behavior in multi-player groups to the response to situations where the player does not know the number of players.