Which Gestalt principle is at work in the old phrase birds of a feather flock together?

Welcome to part 2 of my Nerd Out series, where I take some interesting scholarly research from human-computer interaction & psychology and present them in a simple summary — with notes on how to apply the findings to your chatbot. First part here.

Basic Question:

You have probably heard the proverb that opposites attract. Well that may work for magnets. Others say that birds of a feather flock together. So, which is it — similarities or differences and how can your bot leverage this idea for fun and profit?

We are talking ‘personality’ for your chatbot. Others have written about this before (1,2,3). This is my academic spin on it. Understand that there are many different ways to measure personality. Here are just some of the validated instruments and tests that psychologists use –

But you are trying to build a bot here — not get your PhD. So, I would recommend you use “The Big Five” model. You can remember it’s parts through the acronym O.C.E.A.N as per the image below. Today we are looking at extroversion / introversion and its conceptual relative — dominance / submission.

Basic Answer:

Similarity attraction works. “Birds of a feather stick together” is the angle you want to adopt. Did you know that people with similar personalities are more likely to be happy in their marriage (Berscheid & Walster 1969; Byrne 1971)? It seems pretty obvious, but hey — this is what us nerds like us want to prove.

Now to prove it to you. I have 2 nice experiments to share. Similarity attraction is the theory to use, because it has been applied to human-computer interaction.

EXPERIMENT ONE:

Moon, Y & Nass, C. 1996. How real are computer personalities? Psychological responses to personality types in human-computer interaction. Communication Research, 23(6). Pg. 651–674.

The researchers took a bunch of students from Stanford University and measured their personality to determine if they were ‘dominant’ or ‘submissive’. Then they divided them into 2 groups based on their scores. The researchers took a computer program, more of a game really, where the human uses the computer to help them plan surviving alone in the desert. The user can only carry so much weight, so they have to prioritize what tools to take with them and what to leave behind. The computer is their partner in this decision-making exercise.

But there are actually 2 versions of the computer program — a dominant one and a submissive one:

• The dominant program makes statements like this “the compass should be lower in ranking. You can determine directions by observing shadow movements over a period of time”. And it rates is confidence in this statement as 8/10.
• The submissive program instead says “maybe the compass should be lower in ranking. You can probably determine directions by observing shadow movements over a period of time” and rates its confidence as 3 / 10.

I have highlighted the differences. So you have 4 groups, 4 pairings of human and machine as per image:

And just as similarity attraction theory predicts, the 2 groups where the human personality matched the machines personality, users rated the machine higher for its (i) social attraction, (ii) intellectual attraction, (iii) utility and (iv) emotional satisfaction. All things you want to maximize in your own chatbot right?

EXPERIMENT 2:

Nass, C & Lee, K. 2001. Does computer synthesized speech manifest personality? Experimental tests of recognition, similarity-attraction and consistency attraction. Journal of Experimental Psychology: Applied, 7(3). Pg. 171–181.

Here is a similar experiment, but with a few important changes:

1./ Tested actual vocal cues instead of text. Important for anyone wanting to implement voice at some point e.g. Siri & Cortana or a GPS etc.

2./ Instead of selecting equipment for desert survival, this experiment asked users to listen to a book review and decide if they wanted to buy that book. This is excellent for bot builders — testing a commercial outcome. I assume most readers are at least partially motivated by financial gain.

3./ Tested introverts versus extroverts. These variables are conceptually similar to dominance and submission, but different enough to be important and worth testing.

Here is a summary of the conditions, findings and how the extrovert / introvert variables were sensationalized using vocal cues.

Why does similarity attraction work?

There are dominant theoretical mechanisms (read: explanations) which are kind of related.

1./ Byrnes Reinforcement Model (Byrne 1971): Reality is subjective. As such, we like ‘similar’ personalities because they reinforce the logic and consistency of our own internal model of the world. When we interact with people who are like us and agree with us, it validates our beliefs — and this feels good. TLDR; “This person functions in a logical manner”.

2./ Information Processing Model (Ajzen 1974): We form opinions of people based on the information and cues we have available. This includes information we receive before we even meet them. This theory pus more weight on attitudes. We believe our attitudes are correct, so when we hear someone else has similar attitudes, we draw an inference from that to make predictions about other positive personality traits. TLDR; “I like recycling. They like recycling. Therefore, they are environmentally conscious, just like I am”.

So what? Recommendations:

1./ Don’t just test your bots — test your users! You can get a bunch of valid and proven instruments to measure personality types online. Collect data on your users and build a profile of the average personality. Then you can adjust your bot to suit. **Just make sure you are using a good instrument from an academic journal — DO NOT use some Facebook trash. Email me if you want help finding the right tool.

2./ Read about the big 5 dimensions and start brainstorming. I only covered 1 dimension here. Where do your users sit regarding openness (curious versus consistent)? High openness scorers will appreciate an arty style, more emotional content, unusual ideas, something that stimulates their sense of curiosity. But openness could be perceived as a lack of focus to the low openness (consistent) user. These people want a facts and data driven interaction. This may seem too hard right now, but:

Imagine a future where this personality data is made available to you from Facebook as soon as your user connects. Your bot could adjust its personality on the spot! You be naive to think people are not working on this right now.

3./ Think of personality as more than just the bots name or avatar! Personality will inform the way it answers questions, the firmness of its recommendations, the pauses it takes between responses, the length of its responses — and even more if you plan on adding voice capabilities.

4./ Understand the mediating role of ‘trust’. Ok, just one more piece of theory. This is the latest on the subject. Singh, Tay & Sankaran 2017 tested the model — but actually found that trust mediated the relationship. 2 people could agree about everything, but if A distrusts B — forget about it.

Thank you for reading. I really try to keep my writing grounded in evidence! If this article helped in any way, please hit the clap button below. I will be posting summaries and findings of chatbot related research here, so see you again soon. Let’s connect on LinkedIn!

Which of the following are Gestalt principles of organization?

Which theory of pitch states that pitch is related to the speed of vibrations in the basilar membrane?

Which type of sensory receptors are involved in color vision?

Which of the following is the best description of the vestibular senses quizlet?