Interest in Conversation Requests by Strangers of the Opposite Gender

Summary - The theory being tested was that men are more interested than women in conversation requests made by a stranger of the opposite gender. The actual hypothesis was that men would, on average, show a higher interest (score) than women in talking to a stranger of the opposite gender that initiates a conversation on an Internet chat program. The results were statistically significant in support of the hypothesis, with women averaging a score of 4.69 and men averaging a score of 6.11 on a scale of 1 to 10, with 10 being the highest interest level. Continue reading below, or click here to skip directly to the results section that gives more interesting details.

If you participated in this survey, I'd like to again thank you for your responses. Your screen names and any other personal information have been deleted from my computer. Should you have any questions or concerns, please send an email to psychemail@henage.net. The text that follows was submitted as my assignment.

1. Theory and Hypothesis

My theory is that men are more interested than women in conversation requests made by a stranger of the opposite gender. This theory is based largely on personal observations of stereotypes of men and women and the way they act and react. Men are generally perceived to be more outgoing in initiating conversation with the opposite gender. Women tend to be more accustomed than men to being approached by a stranger of the opposite gender that is interested in becoming acquainted. My personal observations have led me to believe that women tend to react less enthusiastically than men when approached by a member of the opposite gender. Furthermore, the common belief that men are more interested and persistent in pursuing new relationships with the opposite gender further explains the theory.

2. Experiment and Hypothesis

The experiment involves using an Internet chat program to communicate with over one hundred randomly-selected participants, approximately half male and half female. I initiate a conversation with each participant, claiming to be a member of the opposite gender, and ask how interested they are in speaking with me (a stranger of the opposite gender). The hypothesis is that men will, on average, show a higher interest (score) than women in talking to a stranger of the opposite gender that initiates a conversation on an Internet chat program. The independent variable that is manipulated is gender, where I claim to be a stranger of the gender opposite that of the participant. The dependent variable is the "interest score" given by the participant, with higher scores indicating a higher interest in having a conversation with a member of the opposite gender.

3. Procedures

I created a gender-neutral screen name, chat2003talk, with AOL Instant Messenger (AIM). I performed a search for screen names of both genders, using AIM's "Find a Buddy" feature. Each search returns a list of approximately twenty screen names of the specified gender. I performed multiple searches until I had enough names to create a list of well over one hundred females and one hundred males. This list served as the pool of participants, which was drawn from the population pool (all AIM users with screen names listed in the directory).

In my conversations, I attempted to use the same language with both genders. I maintained a list of the phrases I would use, and usually was able to complete the conversation using only phrases from the list (by copying and pasting them into the conversation window). The following paragraph has each of the phrases I used, highlighted in blue text to improve legibility.

I began by sending the message "hi" to several participants on my list. If they did not respond within a few minutes, I would write "hello? "If they still did not respond (or logged off), I would terminate the conversation and discard the name from my list. If they responded with any message, I would then send them a conversation request. If they were female, I would assume the role of a male and write, "I'm a guy from California. Do you want to talk? " If they were male, I would assume the role of a female and write, "I'm a girl from California. Do you want to talk? "If they did not respond, I would discard their name from the list. (If they responded negatively and then did not answer any further questions, I would record their answer as "no, don't want to talk.") Whether they responded positively or negatively, I would then ask, "How interested are you in talking to me, on a scale of 1 to 10 (with 10 being the most)? I'm just curious. " If they would not answer the question, I would try to encourage them by sending one or more of the following three messages: 1, "I just wanted to know. "; 2, "What is your best guess? Please tell me! :) "; and 3, "I have to go, unless you can give me a number! :) " When they gave me a number (or it was clear that they would not give me one), I recorded the result as the "interest score." Upon completion of each conversation, I sent the following message: "Thank you for your response. This was part of a psychology experiment for a school assignment. I am testing the relationship between gender and responsiveness to Internet chat conversations. If you would like to see the results, please go to http://shattern. com/psych. html on Tuesday, February 18th. If you have any questions or concerns, please email me at psych@henage. net. (Don't worry - I'm not keeping records of your screen name or anything like that. ) Thank you. " I would also answer any questions they had at that time.

For each participant, three entries were recorded. The first was gender. The second was whether or not they were willing to talk (in response to my first question: "Do you want to talk?"). The third entry was their "interest score."If they refused to give a number, I recorded their refusal (and didn't use their response in calculations). If they stated that they did not want to talk, and then didn't give a numerical score, then I assigned the lowest possible numerical score of one. Only the gender (the independent variable) and the numerical "interest score" (dependent variable) were used in calculations.

Several adjustments to the "interest scores" were made for the data set used in final calculations. Fractional or decimal scores were rounded to the nearest integer. The highest score from a range of reported scores was used (if the participant responded "6 or 7, " for example, the score of 7 was used for calculations). If participants gave a number above ten or below one, then I recorded the score they gave, but only used the closest number in the range of one to ten for calculations (to avoid, for example, having the score of negative 10,000,000,000 contributing to the average!).

4. Results

Conversations were attempted with approximately 300 participants over the course of about six hours. Scores from 49 women and 54 men were used in calculations. The mean score for women was 4.69, with a standard deviation of 2.89. The mean score for men was 6.11 with a standard deviation of 3.33. The median score for women was 5 and the median score for men was 7. The lowest and highest scores from both groups were one and ten, respectively. [The unadjusted lowest and highest scores for women were negative 10,000,000,000 and 11, respectively. The unadjusted lowest and highest scores for men were "negative infinity" and 10, respectively. ]

From these data it is clear that in this experiment women, on average, have a lower score than men. In order to assess the statistical significance of these results, a one-tailed t-test was performed. The t-value is 2.31, given the 1.42 difference in the mean scores of men and women. The t-value means that there is about a 1.1% chance that the there is no real difference in the mean scores of men and women in the population pool, and it was only chance that the sample pool happened to show a difference in the means. In other terms, we are at least 95% confident that the average woman would score lower than the average man in the population pool. This statistical significance supports the hypothesis that men will, on average, show a higher "interest score" than women in talking to a stranger of the opposite gender that initiates a conversation on an Internet chat program.

5. Consideration of Flaws and Alternate Explanations

Statistical significance from this experiment supports the hypothesis, but does not prove it. In order to increase confidence in the hypothesis, other experimenters using different procedures could perform new experiments. Furthermore, a supported hypothesis does not prove a theory, but does lend support to it. The ability of a hypothesis to support a theory depends partially on the probability that the theory is true (statistical significance) and partially on the degree to which the hypothesis reflects the theory. In this case, I feel that the hypothesis is narrower than the theory, but does do a good job of testing the theory.

There are numerous potential flaws in the way this experiment was performed. Although attempts were made to minimize problems that could cause a significant distortion of the data, some flaws could potentially remain, including (but certainly not limited to) : participants' reflection of the actual population, honesty of participants, experimenter adjustments to the data, and miscommunication. In each of these cases, if a real flaw does exist, it may provide an alternate explanation as to why the results occurred as they did.

Participants' reflection of the actual population – The population referred to in the theory includes all men and women, but clearly the experiment could not pull a random sampling from all men and women. This causes the potentially largest "flaw" in my data and the most probable alternative explanation to the data results – perhaps these results do not apply to the general population. There were several restrictions to the population pool at my disposal. The population pool for this experiment consisted of men and women using AIM during the time period in which I took measurements. Perhaps performing the experiment at different times of the day or different days of the week would produce different results. For example, a sociable person might be out at night with friends, while an introverted person might be more likely to be at home on the computer. Also, all of the participants were somehow listed in searchable AIM directories, whether intentional or not. The type of person that allows himself/herself to be listed in an Internet chat directory probably does not reflect the "average person" (however, it appeared that most were listed unintentionally, which means they may more closely reflect the general population). Finally, the group of people using Internet chat programs does not perfectly reflect the average population – most people do not use chat programs. The results from this experiment could support the theory for Internet chat users, but does not necessarily hold true for the general population.

Honesty of participants – Any of the chat users could easily lie, and it would be difficult for me to distinguish between those that are honest and those that are dishonest. Participants could have recorded a false name and gender when they originally created their AIM screen names. Participants could also give dishonest "interest scores." The potential dishonesty of the participants probably had little effect on the score results, (especially if the dishonest tendencies are somewhat divided between men and women). Part of the reason why I performed the experiment on an Internet chat program is that Internet chat users are potentially more honest and bold in their language and responses because they can usually hide behind their anonymity and are not embarrassed to give their real answers to questions.

Experimenter adjustments to the data – One of the major decisions I faced was what to do with participants that only partially responded to my questions. I decided to discard the results for those that expressed interest in a conversation, but could not produce an "interest score." I discarded these results because it would be inaccurate for me to estimate their interest score, which could be anywhere from one to ten. This probably had little effect on the data, as less than 10% of the final scores fell into this category, and they were almost evenly split between men and women (four and five, respectively). Furthermore, I assigned the score of one to the participants that were not interested in having a conversation and would not give an interest score. I did so because I feel that their refusal of a conversation constitutes the lowest possible score. This had little effect on the final calculations - without this adjustment, the t-score would have actually increased to 2.32, which is still statistically significant.

Miscommunication - I purposely used language that sounded casual in order to simulate a realistic conversation. This may have created confusion. For example, "Do you want to talk?" could be interpreted by some to mean, "Do you want to talk to anyone?" whereas the intended meaning was "Do you want to talk to me?" However, I feel that the language I used probably did not cause a significant error because it is reasonably clear. Also, I occasionally had to sway from my preset list of phrases when participants asked me for my name or age (as I wanted to avoid the influence my reported age or name might have on the experiment). My typical response was " I'll tell you after you answer my question. " I kept any conversation outside my list of preset phrases to an absolute minimum.

The main reason I chose the Internet as a medium is because it would be difficult to have equally balanced independent variables in real life. Attempting to maintain equality between the male and female experimenters in language, dress, age, physical attraction, tone of voice, and personality would most likely prove more problematic and detrimental to the data than would the use of a chat program where these factors play a much smaller role.

6. Future Research

Additional research on this theory should also be extended to "real-life" scenarios. This might include a team of male and female experimenters that approach strangers of the opposite gender on the street and request their name or phone number. I believe that men are more likely than women, for example, to give their phone number to a stranger of the opposite gender. Similar experiments could be performed over the telephone.

Future research on this topic could benefit from the use of computer programs (often called "robots" or "bots") to initiate, carry out, and conclude short conversations on Internet chat programs. This would prove beneficial for two reasons. First, it would reduce experimenter bias by equalizing some small factors such as the time it takes to respond to messages. Second, it would facilitate the ability to survey hundreds of thousands of participants with little effort on the experimenter's side. However, the use of such "bots" is likely a violation of Internet chat program policies and also may cause a nuisance to Internet chat users (although most participants I spoke to were glad to have taken part in the experiment).





The following is a made-up example of the typical conversation that would occur.

chat2003talk:hi

female_chatter: hey, who is this?

chat2003talk:I'm a guy from California. Do you want to talk?

female_chatter: sure

chat2003talk:How interested are you in talking to me, on a scale of 1 to 10 (with 10 being the most)? I'm just curious.

female_chatter: how should i know? i don't even know you

chat2003talk:What is your best guess? Please tell me! :)

female_chatter: ok, maybe about a 6

chat2003talk:Thank you for your response. This was part of a psychology experiment for a school assignment. I am testing the relationship between gender and responsiveness to Internet chat conversations. If you would like to see the results, please go to http://shattern.com/psych.html on Tuesday, February 18th. If you have any questions or concerns, please email me at psych@henage.net. (Don't worry - I'm not keeping records of your screen name or anything like that. ) Thank you.

female_chatter: oh, that's cool



Last update: 20 March 2004.