Brief Problem Statement

What creates a feeling of companionship with a non-player character? How might designers better ensure that a character will resonate with players and create some level of attachment to that character and by extension, the game?

A brief statement of the group’s solutions
to those problems

In order to better understand how to create feelings of companionship, we examined different elements of companionship. Five categories of affect emerged that could be tested against and scored to give a designer a sense for how strong the feelings of companionship a character might elicit from players might be.

These categories are:

• Personhood
• Bonding
• Investment
• Value
• Stakes

Within each category, we propose techniques that may be useful for game designers when creating a companion. These techniques might increase or decrease feelings towards their category and should be balanced as such. Obtaining a high score in all categories does not appear to be necessary for a companion to have a high feeling of companionship amongst players.

Expanded problem statement

Many games include AI companions, friendly NPCs the player has an ongoing relationship with, or NPCs the player can recruit and control directly. Ideally, these companions can create attachment in players that would further endear them to the game through feelings of companionship. Elements that are designed into the nature of these characters might increase or decrease that sense of companionship. It is important for designers to have awareness of how their design might result in terms of companionship.

Expanded solution description

CATEGORIES OF FACTORS THAT ENCOURAGE
FEELINGS OF COMPANIONSHIP

We have categorized aspects of well known character companions from video games, ending up with five categories that factor into establishing feelings of companionship. These categories were tested with a public survey.

Defining each category with example techniques of increasers and detractors.

• Personhood - The overall impression that the NPC is a real person – with their own motivations, autonomy, and internal emotional life
• Theory: if an NPC seems like a real person, players will attach more importance to them
• Factors and techniques for a higher sense of personhood
• Noticeable, unique personality traits
• Don’t let the character overstay their welcome or become annoying
• Features that suggest a life beyond their journey with the player
• Implied detailed lives or backstories
• Motivations & objectives of their own, the illusion of autonomy
• Character has the ability to interact with the environment in natural-seeming ways
• Factors and techniques for a lower sense of personhood
• Ability to take direct control of the character(s)
• Players ability to rename other human NPCs unless it is their child
• Non-human, non-animal, non-anthropomorphized beings
• Breaking the illusion of reality through gamey features or buggy behavior
• Talking about UI menus & buttons
• Bonding - Shared experiences build a deep sense of connection
• Theory: if players go through memorable experiences with an NPC they will build a sense of camaraderie
• Factors and techniques for high bonding
• Shared key experiences (e.g. trauma)
• Romance and/or friendship
• Vulnerability & trust - especially if this is tested
• Bond through common adversary
• Drive relationship via player choice
• Factors and techniques for low bonding
• Undermining the player in story or gameplay
• Superficial writing, character doesn’t have depth
• Unlikeable characterization
• Unappealing performance through AI behavior or voice over
• Repetitive dialogue
• Investment - The player can spend resources (time, attention, currencies) which deepen their engagement with the character relationship
• Theory: expending resources on an NPC will grow a player’s attachment to them, e.g. sunk cost fallacy
• Factors and techniques for high investment
• Complete loyalty quests
• Choose actions to invest in relationship
• Guide ability progression
• Customize appearance
• Protection opportunities through gameplay mechanics or progression (e.g. gearing)
• Factors and techniques for low investment
• Characters don’t remember your actions
• Game systems are poorly camouflaged
• Disposable
• Lacking progression, the character is static
• Value - Character confers benefit to the player that is otherwise uncommon or unavailable
• Theory: a direct benefit will motivate the player to interact more with an NPC than they may otherwise, giving features in the other categories a chance to kick in
• Factors and techniques for high value
• Direct gameplay benefits
• Special interactivity
• Emergent relationships feel personal
• Character is hard to find, conferring value via scarcity
• Synergistic gameplay with the player or party
• Factors and techniques for low value
• Behaviors that can become liabilities (e.g. bad AI)
• Excessive negative penalties
• Low gameplay benefit
• Stakes - The relationship appears subject to consequences from action or neglect
• Theory: loss aversion will drive players to appreciate an NPC more while they have them, and mourning a fallen NPC cements the importance of those that remain
• Factors and techniques for high stakes
• Persistent grievous injuries
• Character can or does leave you
• If the character dies, they stay dead
• Character can betray you
• Character turns evil
• Factors and techniques for low stakes
• Companions are invulnerable
• Easily replaceable
• No consequences if you fail

IMPACT OF SURVEY RESULTS

Some of the categories ended up predicting feelings of companionship with the NPCs we selected, others didn’t, suggesting that we should trim the list of categories in order to make them more useful as a development tool.

From analysis of our survey data, we’ve rated our categories for how effective they were at causing feelings of companionship.

• Personhood - very strong predictiveness
• Bonding - moderate predictiveness
• Investment - weak predictiveness
• Value - strong predictiveness
• Stakes - strong predictiveness

We feel that the low predictiveness of Investment features may reflect on our choice of games or how we framed the category, but as it stands our survey results do not support Investment as a way to increase feelings of companionship.

The survey results supported our categories well, agreeing with the hypothesis that NPCs that demonstrate Personhood and Value, that the player can Bond with and potentially lose will cause more attachment. However, note that the results also show that NPCs don’t need to hit all categories for players to have a strong feeling of companionship with them. For example, XCOM soldiers score high for Stakes and Value, but low for Personhood, and players get very attached to them (not to the point of feeling strong companionship though). By contrast, Elizabeth scores very high for Personhood but low for Stakes and players also get very attached to her.

Another possible  source of error is the time factor - all of the respondents to our survey are using memories of their game experiences from months or years prior. The general praise and continued success of X-COM and Pokemon (as just two examples) may have influenced and shifted their memories of those game experiences. Testing our five factors against feelings of companionship in a more immediate time frame might yield somewhat different results.

Note that it seems more effective to build best-in-class versions of a features in a couple of these categories instead of a lower bar in features spread across all categories.
 

APPENDIX A: Survey questions

For each of the categories, we developed survey questions to explore if we could evaluate how effective an NPC is at being a companion. We developed the following questions based on these categories:

Personhood

• P1 What did you think of this character? [Strongly dislike --Neither dislike nor like-- Strongly like]
• P2 How important was this character to your journey? [Not at all important -- Somewhat important -- Very important]
• P3 This character had motivations unique to themself/themselves. [Not at all true-- Somewhat true -- Very true]
• P4 This character was...Happy [Not at all -- Somewhat -- Very]
• P5 This character was...Clever  [Not at all -- Somewhat -- Very]
• P6 This character was...Annoying  [Not at all -- Somewhat -- Very]
• P7 This character was…Encouraging  [Not at all -- Somewhat -- Very]
• P8 How attached did you feel to this character? [Not at all -- Somewhat -- Very much]
• P9 How much did you want to know more about this character? [Not at all -- Somewhat -- Very much]
• P10 The content this character provided was mostly… [Informational -- Story-related]
• P11 This character felt more like a ....[Tool -- Person]
• P12 This character’s life/backstory was…[Limited -- Detailed]

Values

• V13 How difficult would the game have been without this character? [Very hard -- Neither hard nor easy -- Very easy]
• V14 How autonomous (taking initiative for themselves) did this character feel? [Not at all -- Somewhat -- Very much]
• V15 This character’s abilities enhanced the game. [Strongly disagree -- Neither disagree nor agree -- Strongly agree]
• V16 How much did this character re-enforce your role in the world? [Not at all -- Somewhat -- Very much]
• V17 This character was more of a …[Hindrance -- Benefit]
• V18 How much attention did you have to pay to this character throughout the game? [Too little attention -- About right attention -- Too much attention]

Stakes

• S19 How much did you engage with this character throughout the game? [Too little engagement -- About right engagement -- Too much engagement]
• S20 Was there ever a conflict of ideas between you and this character? [Never -- Sometimes -- All the time]
• S21 How obligated did you feel towards this character...to interact with them? [Not at all -- Sometimes -- All the time]
• S22 How obligated did you feel towards this character...to please them?  [Not at all -- Sometimes -- All the time]
• S23 How obligated did you feel towards this character...to protect them?  [Not at all -- Sometimes -- All the time]
• S24 Did you ever feel in danger of losing this character permanently? [Never -- Sometimes -- All the time]
• S25 Did this character’s programmed behaviors ever break your experience of them as a companion? [Never -- Sometimes -- All of the time]

Bonding

• B26 I shared key moments with this character. [Strongly disagree -- Neither disagree nor agree -- Strongly agree]
• B27 The relationship I had with this character was…[Not at all meaningful -- Somewhat meaningful -- Very meaningful]
• B28 I trusted this character. [Not at all -- Somewhat -- Very much]
• B29 This character had connections with others in the world other than myself. [Strongly disagree -- Neither disagree nor agree -- Strongly agree]
• B30 This character had opinions about things and/or characters in the world that didn’t always align with my own. [Strongly disagree -- Neither disagree nor agree -- Strongly agree]
• B31 How much did your own choice/actions feel like they influenced who this character was? [Not at all -- Somewhat -- Very much]

Investment

• I32 I developed a meaningful relationship with this character. [Strongly disagree -- Neither disagree nor agree -- Strongly agree]
• I33 Did this character feel…[Disposable -- Irreplaceable]

Overall

• C34 What did you think of this character as a companion? [Strongly liked -- Neither disliked nor liked -- Strongly liked]
• C35 How frustrating was this character as a companion? [Not at all frustrating -- Somewhat frustrating -- Very frustrating]
• C36 The sense of companionship in this character was…[Very weak -- Neither weak nor strong -- Very strong]

Survey Results

We ran the survey via surveygizmo.com from December 6th through December 11, 2018. There were a total of 240 completed surveys used in the following results. While there numerous details to be improved upon (survey logic, consistency of scale, consolidated questions per character, to name a few), we were able to get a general idea of a direction to be taken to evaluate our theories.
The following results are a first pass at understanding which questions could help inform designers of companionship traits in NPCs.

Personhood Question Averages

Value Question Averages

Stakes Question Averages

Bonding Question Averages

Investment Question Averages

Overall Companion Question Averages

Sum of Difference between Ideal and Average Response

Variance from av. Affinity (Personhood)

Variance from av. Affinity (Value)

Variance from av. Affinity (Stakes)

Variance from av. Affinity (Bonding)

Variance from av. Affinity (Investment)

Variance from av. Affinity (Companionship Overall)

Question Variance (Personhood)

Question Variance (Values)

Question Variance (Stakes)

Question Variance (Bonding)

Question Variance (Investment & Companionship Overall)