Decision Making – Strategies For UX Designers
Every day we as UX and Interaction Designers ask users of our products to make many decisions and advance through sophisticated designs to unlock their hidden features, rewards, paid goods or free services. Designing interfaces to support users in making the right choices and to elevate them onto the next step in a multi-step journey is not always an easy task. The following article will help you understand how our brain makes choices, estimates probabilities, combines probabilities, and values and assesses risk dimensions.
“The study of how people search for information provides evidence about their decision making strategies.”
What is decision making
Making decisions usually requires evaluating – under varying circumstances – at least two alternatives that differ on a number of attributes. For instance; you are on eBay and need to decide wether you are placing your next bid. Or, you are about to choose a new energy provider or cell-phone plan on a comparison website. Take the example of the person who needed to choose the correct item from a dropdown menu while fiddling with Hawaii’s Airstrike Alarm System. Selecting an alternative requires you to combine often multiple sources of information to form an overall evaluation for each alternative. The study of how people search for information provides evidence about these decision strategies.
In the next section we will look at four popular decision models that do not consider probabilities, because they assume that a person has all the information (values) needed to make an informed decision. In the second section, we will investigate models where decision makers must consider carefully probabilities or fall into the trap of making risky decisions, because relevant information is not at all or insufficiently available.
There are four models of decision making that mostly differ with respect to how people search for relevant information that helps them make a choice.
Decision model #1: Elimination by aspects
Example: Consider you are looking at digital cameras on Amazon and have only a certain budget to spend. You start eliminating the ones that are over your budget. By continuing to select attributes and reject those that do not satisfy some of your minimum criterion, you will soon arrive at your desired item.
The elimination-by-aspects model has the advantage that it does not require any calculations – the cognitive load is low. The decision maker simply selects an attribute (price, colour, feature, etc.) according to some probability that depends on the importance of that attribute. Model #1 and #2 fall both into the basket of non-compensatory models, which are decision strategies that generally reject alternatives that have negative attributes, without considering their positive ones.
Decision model #2: Conjunctive model
Example: Consider you are looking at cameras again, but this time noticed that there are so many cameras available within your budget that you are just simply overwhelmed. You adapt swiftly your decision making strategy by starting to go through each search result and selecting the first camera that satisfies all your minimum criteria.
The conjunctive model is an example of what is known as satisficing search. It is a strategy that follows the conjunctive model and therefore selects the first alternative that satisfies the minimum criteria for each attribute.
Decision model #3: Additive model
Example: Consider you are about to sign up for an online service such as Dropbox for file storage or Spotify for streaming music and are comparing two similar product offerings from two different companies (e.g. Spotify vs. Apple Music, Dropbox vs. Box). You list all your important attributes and then rate each systematically with stars or numbers. You then pick the one that has the higher rating in summary.
One problem with this decision making model is that the rating of the attributes does not account for how the attributes might interact. A single product benefit or attribute from the examples listed above might be so important, that it compensates for a weaker rating of another attribute and vice versa.
Decision model #4: Additive-difference model
This decision strategy is fairly similar to the additive model, but instead of summing up all scores to determine the relative attractiveness of each alternative, you determine the difference for each single score on each attribute. The sum of these differences determines which alternative is more attractive. Model #3 and #4 fall into the basket of compensatory models, which are decision strategies that allow positive attributes to compensate for negative ones.
Concluding the first part; research on how people select a decision strategy has shown that the choice of the strategy depends on task characteristics. People were likely to use a non-compensatory strategy such as elimination by aspects when there were many alternatives to choose from and a compensatory strategy such as the additive model when there were few alternatives. Quite often, decision makers are guided by the dual goals of maximizing accuracy and minimizing effort, causing them to emphasize different strategies in different situations.
Making decisions under conditions of uncertainty seems oftentimes a bit like gambling. It requires that people estimate the probability that a certain event will occur because they do not know which event will occur. Probability estimates are often based on heuristics, which sometimes yield reasonable estimates but often do not.
Example #1: Consider you are being asked to choose between three retirement plans with different payout schemes. It is almost impossible for you to make an informed decision, because at this point you do not have data available that lets you assess your future situation. You might get mislead by media or what someone else suggests. TV tried to influence our decisions many times before.
Example #2: Consider you, as a designer, are being tasked to create and integrate a rating form; you wonder where it should sit in the user journey of your app. We know through availability heuristics that people who are in a happy mood are more likely to recall positive events and a sad mood makes it more likely that we will recall negative events (Blaney, 1986, Source). So your goal could be to find out where your app creates those happy moments, add the call-to-action for your app rating and piggy-bag on these positive moments to receive better ratings from your users.
As we have just seen in the above examples, the availability heuristic proposes that we evaluate the probability of an event by judging the ease with which instances can be recalled. The representativeness heuristics states that the probability of an event is estimated by evaluating how similar it is to the essential properties of its parent population. You can call people who heavily rely on this decision strategy biased (e.g. like goes with like). One problem with basing decisions solely on representativeness is that the decisions ignore other relevant information, such as sample size and prior probability.
Example: Consider you are relying solely on icons or pictorial representations of words when selecting an item in a navigation menu. You’ll likely pick the most representative icon of the visual category that matches your current mental model – respectively it’s the category prototype; looks like X, should be X.
Combining Risky with Evidence-based Decisions
Recall that when we covered the different choice models in the first section of this article, we assigned values (star rating or numbers) to the different dimensions of each alternative in a choice set. It is also important to assign values in risky decision making. Psychologists use a few concepts, such as expected value and subjective expected utility to be able to compare how people make decisions. This will lead us to the topic of risk dimensions.
Most people are likely influenced by the probability of winning, the amount of a win, the probability of losing, and the amount of a loss, but they might not place equal emphasis on these four risk dimensions. The two psychologists, Slovic and Lichtenstein (1968, Source) set up a few gambling experiments and tested the hypothesis that people will be more influenced by some risk dimensions than others. They indeed demonstrated, that participant’s preferences for a few risk dimensions over others had practical consequences.
“As a designer you can influence a users focus on a specific risk dimension by emphasizing a particular dimension. Psychologists refer to this concept as decision frame.”
You can evaluate the effect of decision frames on yourself, by looking at the two experiments below. Which one do you think shows the better deal? Select one from each condition (Plan A or Plan B).
Look closely and you will notice that the information presented in the two plans on the left (Phone Plans) are exactly the same, but scaled differently. One emphasizes price and the other dropped calls. As predicted by the psychologists Burson, Larrick and Lynch (2009, Source) who set up the experiment, the majority of test participants selected Plan B in condition 1 and Plan A in condition 2. See results from both experiments here.
When people make choices that are considered risky, it’s often because they do not perceive the choice as risky.
Example: Consider you are playing a video game and are on a winning strike. You believe that everything is going extremely well and you are in a flow. The game offers you a new challenge that will save you tons of time and rapidly advance you to the next level. All this comes at a high risk; you can loose two levels if you fail the challenge. How do you decide? You see your friend just accepted the challenge. Will you follow? Use the four risk dimensions to calculate your decisions.
Perceptions of risks and risk dimensions are constructed and vary across individuals and cultures. Psychologists demonstrated in a simple experiment with participants from Germany, USA, China and Poland, that the Chinese respondents were more likely than others to show preferences for risky financial options (E.U. Weber & Hsee 1998, Source). So the question to ask here is not “does culture matter”? The right question to ask is “when does culture matter”? Keep this in mind when designing decision trees with the intend to funnel through different users from different cultures.
Example: Consider you are designing a trading app or an online investment management platform. Include a dynamic system that allows people to assess their risk on the basis of their cultural background. Provide smart tools that offer decision aids for different audiences.
In this concluding section, we look at a few more examples of decision research and how findings have been applied to real world situations. We have already encountered examples of heuristics such as availability and representativeness in the work of Kahnemann and Tversky (1979, Source). They both are a bit insufficient in terms of helping people make good judgements. There are better heuristics such as the frequency, imitate-the-majority or the recognition heuristic that allow people to improve their decisions.
Recognition and imitate-the-majority heuristic
A good example for an imitate-the-majority heuristic is this one:
Example: Imagine, you are in the outskirts of an unfamiliar city and need to go into the city center. You are standing on the train platform and recognize a larger group of people on the other side of the train tracks. It’s early morning and something just says: you are on the wrong side of the tracks. If you want to head into the city center, just imitate and follow the crowd.
Can you imagine a good example of a recognition heuristic? Give it some thoughts, until you continue reading.
Example: Imagine, your friend asked you which city has more inhabitants: Guangzhou or Shanghai? You would probably answer Shanghai, since you never heard about Guangzhou.
You can easily apply these two heuristics to your own designs, whether your users need to choose between two payment options, energy tariff plans or three large buttons. Make alternatives recognizable and distinguishable. Keep pictorial- and text-based information aligned with people’s expectations and mental models, so they can make confident choices. If they are still uncertain, allow them to imitate the crowd by providing hints such as; 7 out of 10 people who live in your area and have a similar energy consumption choose energy plan B.
Story model and action-based decision making
There are two more decision strategies to cover. The first one is the story model and it aligns to people’s needs to view a coherent world. Hence why coherent stories presented in a court room hearing are being judged more plausible (Pennington & Hastie, 1988). So next time you want to tell the truth, make it a great, coherent story.
The last decision strategy I would like to introduce is the recognition-primed decision (RPD, proposed by Klein, 1993, Source). It’s often applied in emergency situations and requires experts. Experts usually have faced many similar situations so they can draw on prior experience to make rapid judgements. They often go with the satisfactory alternative not with the best alternative. This strategy allows them to respond quickly. RPD is a recognition-primed model because of the emphasis it places on situation assessment and recognition of what is occurring.
How can you reflect these two decision strategies in your designs? Help your users understand your product offerings, functionalities and choice options by weaving them into a coherent story. Great interaction design is about developing a meaningful and coherent dialog with your users. If they need to make rapid decisions while using your website or app, then always go with standard user interface components and established usage patters so people can quickly identify all available alternatives and select the right options that meet their needs.
Hi! Did you know I am a freelance UX designer and Researcher? You can hire me for your next research session or product design sprint.
Additional resources used for this article:
- Stephen K. Reed — Cognition, 2013
- James & Kerri Goodwin – Research in Psychology, 2014