Who: Daniel Marx, Game Designer at Osmotic Studios
Before co-founding Osmotic Studios with my two colleagues Mel and Michael I studied Computer Science, worked at Daedalic Entertainment and Beardshaker Games while completing my master’s degree in Time-Based Media. When our conceptual prototype for our study project “GroundPlay” won the German Computer Games Award, an advancement award with relatively large prize money attached, we decided to found our own small studio.
Ever since, we have been working on Orwell, our debut title putting players into the role of investigators of a terrorist attack. Players are given control of the eponymous software tool that allows them to read public and private websites online, overhear their phone calls, read their chats, and spy on their private files. Those documents contain information the player may or may not give to a superior who will act upon them.
What: Impactful decisions in Orwell
On a basic level, Orwell is a mostly text-based narrative game that constantly confronts players with choices of varying moral weight. Unlike a typical interactive novel Orwell does not present players with an explicit decision between a set of juxtaposed options (multiple choice) on how to continue the story or which action to take next. Instead it gives players separate pieces of data (we call them “datachunks”) that appear in documents they have found, which contain information they may add to a suspect’s profile.
Adding datachunks to a profile may unlock related documents, through which the player gets access to more datachunks. So the decision whether or not to extract certain datachunks is the most basic choice players are confronted with, and almost everything they will ever do in the game. It was therefore crucial to make every single one of these micro-decisions feel impactful and cause players to think them through.
We did this by attaching the possibility of consequences for characters and the story as a whole to every datachunk: Once added, the information is provided to the players’ superior, the “adviser”, who might consider a character more or less suspicious. Based on the context of other information he possesses on this and other suspects he might chose to intervene, e.g. by ordering an arrest.
In this article I will describe the iterations and considerations during the development process that shaped the choice-based mechanic and the feeling of impactfulness of decisions in the final game.
Why: Creating a choice-driven game without clear right or wrong
Getting inspired by Papers, Please: The beginning
When we first started working on Orwell, we had a game in mind that was supposed to give players the feeling of spying on someone through online documents and pieces of data while telling a convincing story. We also knew we wanted to make a game about (tough) choices. Being fans of Papers, Please we started at a point where players were required to look at digital documents and then decide whether or not that person might pose a threat – very similar to the inspirational game. But we soon came to the conclusion that just cloning from Papers, Please was not only unoriginal, but more severely not able to capture the feeling we wanted to convey. We needed to replace the core mechanic with something that would directly make players feel like they were investigating the data of other people while at the same time providing an unsettling feeling of giving some of it away to someone else without exactly knowing what the consequences for those persons would be.
The next thing that came to mind was to consider the information contained in documents (i.e. textual statements, images) as objects, and to let the player add them to a profile of the target person. New documents would be discovered one by one in regard to information given away. Control-wise we thought about letting people select and add information just by simply clicking on it at first. Finally we went for an actual movement, performed by dragging and dropping the information object from document to profile, because we felt it was needed to further strengthen the sense of literally handing information over.
Searching for the right way: The first prototype
After implementing these mechanics into a first prototype, we ended up with a playable game in which all text and all pictures could potentially contain information and could be extracted. It soon became apparent that being able to add almost anything would overwhelm and confuse players if we didn’t give them any direction about what to look for and only few things actually provided an effect. So we added tasks to the game asking for very specific information the players’ superiors were looking for: a name, a portrait, a relationship, a hobby, etc.
This, however, posed new problems in itself: players would now always have to get tasked with what to look for, giving away to a degree the surprise of what would be coming up – a minor logic issue, but a major dramaturgic one. Even more so, test players reported the game felt extremely linear with no choice at all, despite actually having multiple options to solve a task. Players simply never became aware whenever there was a second solution for a task, and how could they without any hint? Consequently, Orwell was now a game about searching for answers, but not one about choice. We had to iterate the core game mechanic once again.
Figure 1: Screenshot of the first (German language) prototype. Note the tasks in Josef Langley's profile on the left that must be solved by dragging and dropping pieces of information onto them.
Driving the game by choices: Birth of the datachunk
We moved away from the idea of considering everything as processable information, leaving only some predefined statements as extractable to focus the players’ attention. To not let people guess which items inside documents were usable we had to point out what options there were, which we did by clearly highlighting them. At the same time we got rid of the task system in order to give players a feeling that they were investigating by their own discoveries rather than ticking off a to-do list. Players were now able to pick information they deemed important and add them to a person’s profile in order to possibly uncover new documents. The first datachunks were born.
With datachunks we also preserved the idea of having explicit decisions between multiple options. At first this worked by dragging all datachunks belonging to a conflict into the profile to first juxtapose them and thereby granting the player to compare the options before going for only one of them. But this mechanic was fundamentally undermining the concept of permanence of anything uploaded. So we decided to rework it to work analogue to “normal” datachunks, with the tiny difference of permanently deactivating all conflicted counterparts and warning players whenever they hadn’t found all of them yet. Not to be too restrictive, the system did not prohibit players from making a choice without knowing all the options. In fact, we wanted to actively tempt players to sometimes act prematurely.
Further tests showed that without imposing any drawbacks, players would tend to simply look for anything highlighted (except conflicted datachunks) and drag and drop it into the profile in order to gain more documents and therefore progress, again not making them weigh up their actions. But a game-over mechanic forcing players to start over for giving away something “wrong” or unimportant would be tedious in a highly narrative game. Even worse, imposing a defined sense of “right” or “wrong” on available datachunks would have removed any moral ambiguity from choices, something we felt was extremely important to make choices more intense.
Players’ actions had to have tangible consequences, a thing we had already planned for in terms of a branching story, but since we were a rather small team, it was not possible to have major game changing consequences for any combination of datachunks people would be able to give away. So, to make them question the moral behind their actions, more important than actual branching was to evoke a feeling that anything players gave away might have an impact on the persons they were investigating.
Evoking a sense of consequence: Context-based feedback by the adviser
We came up with the idea of the “adviser.” The adviser is a character who comments on the information the player gives away, and therefore is the central element to provide feedback in Orwell. On a very basic level the adviser makes statements about what has just been uploaded, a line of text to let players know the new information they found has been noted. But his true role goes far beyond that:
Over the course of the game players build up a more and more nuanced profile of information on characters. Since the adviser is supposed to only know what players give away they construct the perspective he has on the suspects, which is of great importance since he is the only one who can actually take action like declaring someone a suspect or ordering an arrest.
Figure 2: The overhauled prototype, now containing the adviser in the upper left corner. Also the tasks have been removed from the target's profile. Instead, extractable information is highlighted in orange in the document on the right.
To contribute to a feeling of continuous evaluation, players should be made aware as often as possible what the adviser was thinking about the target persons. To do this, it was necessary to look at the context of what else could be present in a character’s or even other people’s profiles at the time of a new upload, and to prepare unique statements for significant cases.
On the level design side of development this meant being aware of what combinations of information could be present at any given time, and specifically designing for various viable and interesting perspectives of characters. Technically the statements are selected by checking for those specific combinations of uploaded datachunks, sometimes in regard of the progress players have made so far, and then showing the corresponding text message fit for the specific case, or else a more generic alternative. Because order and episode progress also matter, some datachunks amount to six and more completely different commentaries with additional variants.
Combined with the progression system of having to give information away in order to gain new one in mind, this pitches the players’ interest of driving the story forward by unlocking new documents and helping their case against giving away information about target persons possibly leading to severe consequences for them, creating a moral conundrum.
Figure 3: The finalized UI with the adviser ("Symes") in a modular pop-up window. Players often proved to be oblivious of non-modular approaches positioned not prominently enough on-screen.
Clarifying choices: The datachunk information pop-up
With pieces of information taken out of context there’s always (subjective) interpretation involved, so from the first prototype on one issue constantly whenever people where playing the game, but also when designing levels: It wasn’t always clear what would actually be added to a suspect’s profile when it was being extracted. This was especially true for more ambiguous information apart from hard facts like a suspect’s occupation or name. Although ambiguity was something we wanted in the game, we had to take care of because ultimately a blind choice is a meaningless and potentially frustrating one.
To give players a better idea what to expect when adding a datachunk we added a pop-up window informing players about anything that would be added to a person’s profile when this datachunk would be extracted. This would never reveal how the adviser would interpret the information and therefore show the consequences, though.
We added a disable function to the datachunk information pop-up, allowing to unmark datachunks that seemed not important or wrong so that they would not appear as active any more. The way that story progression in Orwell worked implicated that some of the datachunks had to be mandatory, i.e. they always needed to be extracted. To that end it was important to provide a function allowing to undo disabling datachunks in case the player would have to reconsider later. By adding the disable function we hoped to give players the possibility of consciously deciding not to use a datachunk and flagging it accordingly rather than “doing nothing”, something we thought might feel like having forgotten to act at all.
Figure 4: The datachunk information window allows considering information before it is added to the profile and to disable the datachunk.
Adding layers of moral conflict: System, character, and player bias
With the contents clarified now we realized we could go and create datachunks around even more vague or out-of-context statements. What if the Orwell system would try and parse such information, but would often be mistaken about its true, more complicated meaning? This presented a great opportunity to create an additional layer of conflict and therefore making players think.
To clarify to players that they had to expect datachunks being this kind of ambiguous we placed some of them right at the start of the game. In the first episode we added some datachunks containing undeniably wrong or misleading information, something that would immediately cause a noticeable, undesirable or at least questionable effect. When extracted, the adviser would confront players with the “mistake” they had just made, or some other easily comprehensible minor impact on the story would play out.
A good example for this is the “credit card stolen” datachunk: Within the very first conversation in the game Josef asks Cassandra whether she knows where his credit card is, whereupon Cassandra answers that she “snatched the card from his desk.” From the context of the conversations it’s abundantly clear this is just a playful remark between to persons who are close. The Orwell system though is oblivious to this context, interpreting the statement as datachunk including the information that “Cassandra has stolen a credit card from Josef.” In case players decide to extract this information the adviser Symes, only knowing the “fact” Cassandra stole a credit card leads to him locking the card. This again causes the next conversation the two characters have play out differently to confront players with immediate consequences to their action.
Figure 5: A screenshot of Articy:Draft 2, the tool we used for level design and scripting. In the top image you can see the inner workings of the “credit card stolen” datachunk: Adding this one will result in short-term feedback (branching comments, white nodes) based on another info the player might have added before. The bottom image depicts a later consequence of having added the “credit card stolen” datachunk, causing a different dialogue branch to be selected.
On top of this bias by the Orwell system there’s one more dimension to consider when making a decision in the final game: Information may be biased by characters. Naturally a character does not always state the truth or be very subjective about certain aspects, which is why some datachunks does contain biased information. To further this concept we also turned the adviser into a full-fledged character with motives and goals within the story, therefore leading to a biasedness of his own to consider in regard to consequences. Last but not least, when writing the story for the game we paid attention to make all characters likable in their own regard, so that players might sympathize with at least some of them, possibly leading to being prejudiced about the whole situation themselves.
Creating a game based on decisions without having an in-game punishment or reward mechanic for any of the players’ actions has the dangerous pitfall of feeling hollow and meaningless. To counter this circumstance we put a lot of effort into attaching a sense of meaning and tension to each of the choices.
Is the system we have created perfect? Certainly not. Theoretically, the game can still be exploited rather easily by extracting everything, just as before if players don’t care at all. Was it still worth the effort? Absolutely! For the vast majority of players the effort we put into framing the context of choices seems to work out rather well, as a lot of positive feedback and reviews have testified.
With all the multiple layers in place decisions in Orwell often amount to questions like this: Do you reveal that a character who is a single parent but impulsive is armed based on a vague statement towards a superior already on edge because his job depends on immediately getting hold of them?