“I Am in Your Computer While We Talk to Each Other” a Content Analysis on the Use of Language-Based Strategies by Humans and a Social Chatbot in Initial Human-Chatbot Interactions

Figures & data

Table 1. Intercoder reliability for all variables in phase 1.

Variables	Krippendorff’s alpha	Percent agreement
Phase 1: Chatbot’s utterances
Uncertainty reduction strategy	0.80	0.87
Intimacy of self-disclosure and question asking	0.58	0.66
Self-disclosure reciprocity	0.63	0.89
Question asking reciprocity	0.51	0.91
Similarity	0.46	0.97
Empathy	0.92	0.99
Humor	0.47	0.96
Appropriateness	0.58	0.98

Note. Self-disclosure and question asking were coded as part of the variable “uncertainty reduction strategy” (1 = self-disclosure, 2 = question asking, 3 = other). The variable follow-up questions was part of “question asking reciprocity” (0 = none, 1 = follow-up question, 2 = reciprocal question). The variable similarity also includes dissimilarity (0 = none, 1 = expression of dissimilarity, 2 = expression of similarity).

Table 2. Intercoder reliability for all variables in phase 2.

Variables	Kappa	Percent agreement
Phase 2: Participants’ utterances
Uncertainty reduction strategy	0.77	0.86
Intimacy of self-disclosure and question asking	0.82	0.88
Self-disclosure reciprocity	0.68	0.92
Question asking reciprocity	0.40	0.92
Similarity	0.75	0.99
Empathy	0.64	0.99
Humor	0.65	0.99

Note. Self-disclosure and question asking were coded as part of the variable “uncertainty reduction strategy” (1 = self-disclosure, 2 = question asking, 3 = other). The variable follow-up questions was part of “question asking reciprocity” (0 = none, 1 = follow-up question, 2 = reciprocal question). The variable similarity also includes dissimilarity (0 = none, 1 = expression of dissimilarity, 2 = expression of similarity).

Table 3. Means and standard deviations for all variables for the participants’ and Mitsuku’s utterances

Dependent variable	Participants	Mitsuku
Amount of self-disclosure**	0.25	0.23
	(0.13)	(0.13)
Intimacy of self-disclosure***	1.63	1.20
	(0.32)	(0.67)
Reciprocity of self-disclosure***	0.16	0.09
	(0.11)	(0.09)
Amount of question asking***	0.21	0.28
	(0.11)	(0.11)
Intimacy of question asking	1.95	1.46
	(8.47)	(0.26)
Reciprocity of question asking***	0.02	0.03
	(0.03)	(0.03)
Follow-up questions***	0.04	0.08
	(0.06)	(0.07)
Similarity**	0.01	0.02
	(0.02)	(0.03)
Dissimilarity	0.00	0.00
	(0.01)	(0.01)
Empathy	0.00	0.00
	(0.01)	(0.01)
Humor***	0.01	0.03
	(0.02)	(0.04)

Note. Standard Deviations appear in parentheses below means. **p < 0.01. ***p < 0.001.

Table 4. Means and standard deviations for the Chatbot’s inappropriate responses.

Interaction number	M	SD
1	0.08	0.08
2	0.07	0.07
3	0.08	0.08
4	0.07	0.07
5	0.08	0.08
6	0.07	0.08
7	0.08	0.08
Overall	0.07	0.07

Note. The inappropriateness of the chatbot’s responses was calculated by dividing the frequency of the occurrence of this variable per interaction by the total number of utterances in each interaction.

Figure 1. The decrease of amount of self-disclosure for chatbot Mitsuku over the seven interactions.

Figure 2. The decrease of expressions of similarity for chatbot Mitsuku over the seven interactions.

Figure 3. The relationship between the chatbot’s utterances and the participants’ utterances over time. The raw data points are accompanied by a regression line and corresponding 95% confidence bands. Note. “P stands for “participant,” C stands for “chatbot.”

Figure A1. The relationship between chatbot question asking and human self disclosure over time. The raw data points are accompanied by a regression line and corresponding 95% confidence bands.

Figure A2. The relationship between chatbot question asking and human question asking reciprocity over time. The raw data points are accompanied by a regression line and corresponding 95% confidence bands. Note. “H stands for “human,” C stands for “chatbot.”

Figure A3. The relationship between chatbot question asking depth and human self-disclosure depth over time. The raw data points are accompanied by a regression line and corresponding 95% confidence bands.

Figure A4. The relationship between chatbot self-disclosure and human self-disclosure reciprocity over time. The raw data points are accompanied by a regression line and corresponding 95% confidence bands.

Figure A5. The relationship between chatbot self-disclosure and human self-disclosure depth over time. The raw data points are accompanied by a regression line and corresponding 95% confidence bands.

Figure A6. The relationship between chatbot self-disclosure and human expression of similarity over time. The raw data points are accompanied by a regression line and corresponding 95% confidence bands.

Figure A7. The relationship between chatbot expression of similarity and human expression of similarity over time. The raw data points are accompanied by a regression line and corresponding 95% confidence bands.

Figure A8. The relationship between chatbot human expression of similarity and human self disclosure over time. The raw data points are accompanied by a regression line and corresponding 95% confidence bands.